Engineering in the Ancient World

ENGINEERING IN THE

ANCIENT WORLD

ENGINEERING IN
THE ANCIENT
WORLD

J. G. LANDELS
Unimsity qfReadittg

UNIVERSITY OF CALIFORNIA PRESS
Berkeley & Los Angeles
1978

-rh±s Ono

^^^^'^^'^^OK Copyrighted matferial

UNIVERSITY OF CALIFORNIA PRESS
Berkeley &. Los Angeles

ISBN: 0-520-03429-5
Library of Congress Catalog Card Number: 76-52030

© J.G.Landds 1978

Al! rights reserved. No part of this publication
may be reproduced, stored in a retrieval
system, or transmitted, in any form, or by
any means, dectranic, mechanical, photo-
copying, recording or otherwise, without the
prior permisiion of the publisher.

Printed in Great Britain

Copyrighted material

Contents

1. POWER AND ENERGY SOURCES 9

Man-power 9

Animal power 13

Water power 16

Wind power 26

Steam power 28

2. W ATER S UPPLIES AND E NG INEERI NG M

Appendix — The sizes of measurement nozzles, and

Frnnfiniis' arifhmefir 53

3. WATER PI:MPS 52

4. CRANES AND HOISTS 84

CATAPULTS 99

fi. SHIPS AND SEA TRANSPORT 133

Appendix — Methods of estimating; the

maximum speeds of oared vessels 166

7. LAND TRANSPORT LZO

8. THE PROGRESS OF THRORETTCAL

KNOWT.EDOR IBS

9. THE PPTNCIPAE GREEK AND ROMAN

WRITERS ON TECHNOEOGTCAL
SUBJECTS 199

Hero of Alexandria 199

Vitniviiis 2QS

Frontinus 211

Pliny 215

BTRT.TOOR APHV 21B

INDEX 221

Cc lial

Preface

The purpose of this book is to discuss and illustrate a number of
technological achievements in the Greek and Roman world, all of
which have in common the application of power for engineering
or other similar purposes.

It would be quite impossible in a volume of this size to attempt
a complete review of the subject on the scale of Bliimner's
Technofopie und Teryninologie . . . etc, or the History of Tech-
nology ediled by Singer, Holmyard and Hall. Nor would it be
appropriate to attempt very lengthy and detailed treatments of
individual topics such as those compiled by R. J. Forbes in his
scries Studies in Ancient Technology, Instead, an attempt is made
to give the modem reader (whether a student of classical civiliza-
tion, or a layman interested in the history of science) some insight
into the mechanical skills of the two most fascinating civilizations
of andent Europe.

It may justly be said that most major histories of technology
(including those mentioned above) are subject to two critidsms.
Firstly, the written sources are not always examined in detail, and
the Greek and Latin terminology is not usually analysed. In this
book most of the key words are gi\cn (in English transliteration
where appropriate), so that the non-classicist can see something of
their etymology, and perhaps feel a little closer in thought to the
ancient writers. Secondly, in most standard histories, the archaeo-
logical evidence is treated in a descriptive way, and very little
attempt is made to envisage mechanical contrivances in action. I
have tried to do some calculations on most of the machines and
devices; and although some of them, in the nature of thingis, can
be no more than rov^ guesswork, they may hdp to avoid the more
absurd under-estunates — or over-estimates — of what the devices
could do.

The transladons of passages quoted from Latin and Greek

authors are all my own. Other available versions may well be
superior in literary merit, but even the most eminent of translators
can sometimes make extraordinary mistakes on technical matters;

Copyrighted material

8 ENGINEERING IN THE ANCIENT WORLD

the shade of Odysseus must have derived great amusement from
some of the versiom of his boat-building in Odyssey V.

As a dassidst, I have received during the writing of this work
welcome advice and help £rom a number of Reading University
colleagues in a variety of disciplines. In particular I would wish
to thank Professor J. E. Gordon of the Department of Engineering
and Cybernetics, who has given vahiablc help on two particular
points — the properties of animal sinew (discussed in Chapter 5)
and the speeds attainable by oared vessels (Chapter 6). The graph
on p. 167 is reproduced by his kind permission.

Among many other colleagues who have given kind advice, 1
would particularly mention Dr Douglas Balch and Dr N. K.
Jenkins of the Department of Physiology and Biochemistry, and
Dr M. G. Fulford of the Department of Archaeology.

Dr Malcolm Pctyt gave valuable help at the proof stage and —
not least — my wife Jocdyn has been helpful and encouraging at
all tunes, and capably carried out the traditional spouse's task of
compiling the index.

I also wnh to thank Mr Edgar HoUoway, who prepared all the
illustrations from my very rough sketches.

My former colleague in the Chissics Department, Professor K.
D. White, gave valuable support and encouragement during the
early steiges of the writing, for which I am grateful.

Finally, I must thank Professor M. I. Finley, the Editor of this
series, whose shrewd insight and kindly advice have been invalu-
able throughout the long and exacting task of writing this volume.

J.GJL

Note : Figures are given in both British and metric units. To avoid
confusion the abbreviation of litres is given by L

Copyrighted material

1

Power and energy sources

The sources of power available in classical antiquity were severely
limited by comparison with those of the present day. Virtually all
work was done by man-power or animal power, and the kind of
constraint which this imposed may be seen from a ample illustra-
tion. One gallon of petrol may seem very expensive nowadays, but
if used in an ordinary engine of average efficiency it will do the
equivalent work of about 90 men, or of nine horses of the smallish
size used in the ancient world, for one hour. Water ])Ovvcr was
used for pumping and industrial purposes, but probably not much
before the first century B.C. The theoretical possibilities of steam
power, hot air expansion and windmills were known, but appa-
rcndy never exploited except on a very small scale, and not in use-
ful or practical applications.

MAN-POWER

The most common mode of employing man-power was in the
handling and porterage of small burdens of the order of 20-801b
(9-36kg). This is discussed in detail in Chapter 7, and all that

needs to be done here is to note a very important limitation, which

should be quite obvious, but is all too often forgotten. If a burden
requires more than one man to liandlc it, its size and shape must
be such as to allow the necessarv number of men to stand close
enough and get a grip on it. Tor example, in the fifth century B.C.
the columns of Greek temples were buill up from a number of

sections, called 'column drums'; these might be anything up to 6ft
Gin (2m) in diameter. The only possible place to grip such a lump
of stone is around the lower edge, and it would be very difficult for
more than 18 men to get into position to grip it at once. It follows,
therefore, that if its total weight was more than about a ton (as it
often was), they would be unable to lift it up off the ground, let
alone move it, turn it round or position it on a column. They
might just be able to roll it along level ground on its edge, but that

Copyrighted material

10 ENGINEERING IN THE ANCIENT WORLD

would be all. When, therefore, people say *of course they had
thousands of slaves to do the building for them', two facts should
be remembered. Though the Pharaohs in Egypt may have had
vast resources of manpower, Greek and Roman buildhig contrac-
tors rarely had more than a small labour force, and In any case, no
matter how many they may have assembled for the more ambitious
projects, they could never have maii-liarielleJ the larger stones
used in classical buildings. Either one of the lifting devices des-
cribed in Chapter 4 must have been employed or else the very slow
and exiravaj^ant method of building a ramp, and dragging the
stones up the slope on rollers.

There were two important mechanical devices for harnessing
man-power. One was the capstan or windlass, particularly useful
on cranes or aboard ship. The power could be transmitted over a
distance by ropes, its dkecdon could be changed by pulleys, and
the force could be multiplied by block-and-tadde arrangements.
The windlass itself has a built-in mechanical advantage. It was
also found to be ideal where tracdon was required, of low power
but finely and accurately controlled. Two medical uses illustrate
this. One was the so-called 'bench of Hippocrates* — a plinth with
a windlass at each end to provide the extension needed for re-
ducing fractures and dislocations of the arms and legs. The other
was a device apparently used by some g\'naeco]ogists — a small
capstan mounted below a 'midwifery stool', used for extracting a
foetus from the uterus.*

It is generally agreed that the Greeks and Romans did not,
apparently, discover or use the crank in place of the handspikes on
a windlass. Hero of Alexandria speaks of something called a 'hand-
holder' {cheirolabe) for turning axles. This might have been a
crank, but there is no proof that it was. There was at least one
situation in which the main advantage of the crank could have
been exploited, and where its disadvantage would not have been
noticed — the repeater catapult. Since it was not used on that
weapon, it seems almost certain that it was not known to the
designers.

How serious a drawback was the lack of this device ? The answer
seems to be — rather less than is sometimes suggested. The only
real advantage of using a crank is speed. A single grip (firm, but
loose enough to allow the handle to turn in the palms of the hands)

^Hippocrates, On Joinis chapter 72: Soranus, GjfnMceia XXI, 68.

Copyrighted materia!

POWER AND ENERGY SOURCES U

can be m ai nt ai ne d all the time, whereas vdth handspikes the grip
has to be changed, usually four times per revolution. But in situa-
tions where speed is less important, the crank has a positive dis-
advantage. The force which can be applied to it varies according

to its position in relation to the operator, reaching a minimum
twace during each revolution when the handle crosses a line drawn
through the operator's shoulders and the axis of the crank, and
a maximum when it is roughly at right-angles to that line. This
is why a car starting-handle used to be be so arranged that the
points at which most force is needed to turn it occurred when
the handle was at 'two o'clock' and *eight o'clock'. This imposes a
serious limitation on the crank. When it is under continuous load-
ing (e.g. on a crane when the load is raised, or a well-head when
the bucket is full), the reverse thrust applied to the crank handle
by the load must never exceed the minimum applied by the
operator at the two weakest points of the cycle. If it does, the
handle will fly backwards, and once it has started swinging round
the load may acquire momentum and make the handle impossible
to stop. To avoid this danger, most modern cranked winches are
fitted with a ratchet. Such a device, dating from the late fifth
century B.C., was found near some naval installations at Sunium,
and may have been used on a winch for hauling ships up slipways.

The implications for ancient devices worked by handspikes are
dear enough. For cranes or hoists of any kind the use of a crank
would have lowered the handling capacity by some 20-30%, and
it seems rather improbable that a slight increase of speed would
justify that sacrifice. On the repeater catapult the slider was fitted
with pawls and a ratchet, and would only fly forward a short
distance if the tension on the draw-back cord were relaxed. It
would therefore have been reasonably safe to use a crank on the
capstan at the rear of the machine and thereby speed up the load-
ing operation — a particularly important benefit, for that parti-
cular weapon.

The other mechanical device was the treadmill — a pair of
vertical wheels with treads (like those of a step-ladder) between
them. It has become very difficult nowadays to talk, or even to
think about this apparatus unemotionally, and in purely engineer-
ing terms, but in fact, if well designed, it can be one of the most
efficient devices for this piupose, and the most comfortable for the
operator — in so far as any continuous, monotonous physical wcnk

Copyriyhioa inaici lal

12

ENGINEERING IN THE ANCIENT WORLD

can be comfortable. The basic action is not unlike that of pedalling
a Hcyde, and it is significant that recent attempts to reach the
absolute limits of the human body's capabilities, in the develop-
ment of man-powered flight, have mosdy used that airangement.
The difference is that a cyclist pulls on the handlebars, and uses
the abdominal muscles as nveU as the leg mucks; the treadmill
operator uses the reacdon from lifting his body weight, mainly
with the leg muscles.

A very useful feature of the treadmill, especially when used on
a crane, is that the torque, which determines the pull on the hoist-
ing-cable, can be easily and accurately adjusted by the operator
shifting his position on the wheel. The maximum torque is obtained
when the operator treads the wheel at a point on a level with the
axle (this can only be done from the outside). If he treads above
that point (outside) or below it (inside) the torque is less, and if he
stands direcdy above or below the axle it is zero. Thus the amount
of torque required between the maximum and zero, can be
obtained by moving forwards or backwards.

This may possibly afford an explanation of a rather mysterious
length of wood with notches along one side, found near the Roman
water pumps in the Rio Tinto mines. When these pumps (which
themselves acted as treadmills) were being used in a series, it would
be very important to keep the output of each of them constant,
and consistently the same as that of the pumps above and below
— otherwise the simips would either empty or overflow. If this
piece of wood was one of two beams supporting a movable hand-
rail, the necessary adjustments for men of different weights work-
ing the same pumps at different tunes could be made by shifting
the rail along oac or two notches, forwards to reduce output or
backwards to increase it.

A second valuable feature of the man-powered treadmill is its
mobility. The crane shown on the monument of the Haterii (p. 84)
could presumably have been dismantled, and its jib laid horizon-
tally on one or more carts, while the treadmill itself could have
been rolled along any reasonably level road (that was also one
method used for transporting column-drums). There was, in fact,
no other suitable power source available. Wind power is hope-
lessly unreliable, and a builder would be extremely lucky to have
water power available on the site at all, let alone near enough to
any particular building. A glance at the later history of cranes

Copyrighted material

POWER AND ENERGY SOURCES

13

shows that the treadmill continued to serve this need right through
the Middle Ages and Renaissance, and that the first alternative to
be made effectively mobile was steam, as used on railway break-
down cranes. Indeed, the problem is still with us. Owing to diffi-
culties of gearing and transmission the internal combustion engine
is not very suitable for large cranes, and the cost of laying supply
cables makes it uneconomical to use electricity for anything less
than a large and lengthy building project.

The Greeks and Romans also used manpower for the propul-
sion of virtually all fighting ships. Merchant sliips, except for
quite small ones, were normally under sail. Warships used sails
on long voyages, or while cruising on patrol, but in battle condi-
tions, or during a battle alert, they usually left mast, yard and
mainsail ashore, to cut down weight to the absolute minimum,
and relied entirely on rowers.

ANIMAL POWER

From remote antiquity there has been a contrast between the
worldly animals used in the Mediterranean area and those used
in northern Europe. The predominance of the horse in northern
Europe, closely related to climatic and ecological factors, could
never have occurred in classical Ckeece, and did not affect Roman
practice to any great extent except in so far as Roman armies
came into contact with the peoples of France, Germany and central
Europe. The situation in classical Greece is summed up both accu-
rately and poetically by Aeschylus in a passage of his Prometheus
Bound. The hero, describing his services to mankind, says (lines
462-6)

'And I was the first to link oxen beneath the yoke
With yoke-straps, to be man's slaves, and with their bodies'
strength

Give him relief from the heaviest of his toil;

And to the chariot-pole I brought

Horses that love the guiding reins.

Delight and pride ci massive wealth and luxury*.

The slowness and ugliness of oxen (a generic word, meaning *great
knobbly beasdes' is used in the Greek original) is contrasted with
the speed and elegance of horses. The assertion in the last line,

14 ENGINEERING IN THE ANCIENT WORLD

that horses were expensive to buy and maintain, is borne out by
the fact that several words denoting social and economic status
in classical Greece were connected with horses. The word hippeus^
referring to a particular income-group, originally meant a man
wealthy enough to own his own horse and (in wartime) to fight in
the cava]ry of the dtizen army. In Athens the next lower property-
clas^cation was zeugites, meaning a man who owned a pair of
oxen. The historian Hci odotus, wishing to stress the great wealth
of a particular family, calls them tethrippotrophon — able to main-
tain a four-horse racing chariot (for entry at the races during the
great games at Olynipia, Delphi and elsewhere). The 'conspicuous
consumption' of such a family must have made a deep impression.

By contrast, a pair of oxen could be fed much more cheaply, on
inferior fodder of a kind available in areas of Greece and Italy
where the pasture was not adequate to support horses. They
yielded a return on the owner's mvestment; they could pull a
heavier load than two horses of comparable size. Their progress
was slower, but then speed was not the most important considera-
tion in ancient farming or transport. Farm animals had to be itA
all the time, whether in use or not; a transport contractor would
naturally want to complete each job as soon as possible to be ready
for the next. But to use horses to speed up his operations would
have been quite impractical. And finally — an important point for
people living close to subsistence level — when their working life
was over, oxen could serve as food. The meat would be tough as
old boot, no doubt, and would need a long spell in the stewpot,
but it would be better than nothing. The Greeks and Romans, for
reasons not clearly defined but presumably religious, did not as a
rule eat horsemeat.

The one advantage that the horse had over the ox was speed,
and it was precisely in those situations where speed outweighed
everything else that the horse was used — in warfare and in chariot-
racing. The high mobility of the cavalr)' gave that arm its parti-
cular role in battle tactics, and on the race-course a chariot, made
as light as possible, and drawn by a matched team of two or four
horses, represented the ultimate in speed to the Greeks from the
eighth century b.c. onwards, and to the Romans after them.

Oxen, then, propelled the heavy lorries of the ancient world,
and highly-bred horses its Aston Martins and its Lamborghinis.
Between these extremes of utility and luxury came the small

Copyrighted material

POWER AND ENERGY SOURCES

15

traveUing vehicle for passengers or light merchandise, drawn by
donkeys or mules. These axidmals could move rather faster than
oxen, but not as fast as horses. They cost a little more to feed (in
proporticm to their weight and puUing capacity) than oxen, but
only about 60-70% of the cost of horses.

The use of animals in transport, and the problems connected
with harness, are discussed in Chapter 7. Apart from transport, the
use of animal power was rare. In mining operations it seems to have
been almost negligible, for obvious practical reasons. Unless there
was access via horizontal tunnels ('adits'), it would be very difficult
indeed to get animals into or out of a mine, and ancient workings
did not normally include entrance adits or any galleries or spaces
in which animals could be kept, fed and housed imderground. The
haulage of ores and spoil seems to have been done exclusively by
man-power, using buckets on ropes, and it was extracted via the
nearest shaft, not taken along any great distance underground.

Until about the first centurv b.c. animals were not used in mill-
ing. The only type of mill which can be operated by a horse or
donkey is a rotary mill, and that invention did not come into the
classical Greek world at all. The so-called Pompeian mill, with a
fixed lower stone of conical shape, and a rotating upper stone
shaped like an hour-glass was c^uite certainly designed to be turned
by animal power, despite the fact that the space available in some
of the buildings for the animals to walk roimd seems very limited
indeed. The earlier 'pushing' type of mill, in which a grinding
stone is pushed back and forth over a trough, must have depended
on human effort. Such work was sometimes imposed on slaves as
a punishment, but at all times it had to be done by someone, and
as a punishment it was probably not much more severe than the
*spud-bashing' to which army offenders used to be subjected — a
tedious, irksome job which nobody would do from choice. Some
illustrations of rotary mills being turned by horses give a highly
idealized picture of noble steeds striding around; in real life, the
oldest and most broken-down horses and donkeys were put to this
Jdnd of work — the last stage on the road to the knacker's yard.

Finally, there is a bizarre invention described in a Latin work
written in the latter half of the fourth century a.d., but almost cer-
tainly never constructed. The author's name is not known, and the
work is usually referred to as Anonymus De Rebus Bellids, Oxen
are used to propel a ship (Fig. 1). They walk around in pain, at

16 ENGINEERING IN THE ANCIENT WORLD

opposite ends of a capstan-pole on a vertical axle. Through a
gearing system (not described, but clearly a crown wheel and
pinion, as used in water-mills) thb axle drives a horizontal one
athwart the ship, with a paddle wheel on each end — the descrip-
tion of the paddles has some verbal resemblances to Vitruvius'
description of an undershot water wheel (X, 5, 1.) We are not told
whether the paddle-wheel shaft was higher or lower than the
platform on which the oxen walked, but since they were *in the

Fig. 1 . Oxen used to propel a ship.

hold of the ship', it seems more likely to have been the former. The
total number of oxen is not specified, except that there was more
than one pair. Though there is no theoretical reason why this
should not work, the whole idea does not sound very practical.
The space needed for the oxen to move around would be con-
siderable — a circle of 10ft (3m) diameter at the very least. If we
assume three capstans, the ship would require a beam of about 13 ft
(4 m) and a length overall of at least 43 ft (13 m), and at 'six ox-
power' such a vessel would be rather under-engined. Communica-
tion between the 'bridge' and the 'engine-room' might also be a
trifle difficult.

WATER POWER

Early Greek poetry contains striking passages in which the destruc-
tive force of rushing water is used as a piece of telling imagery, but
the problems of harnessing such power and using it to drive machi-

POWER AND ENERGY SOURCES

17

nery were apparently not explored until the early part of the first
century b.g. According to the geographer Strabo PQI, 3» 40) a
water-mill was built in the kingdom of Mithiidates» at Kabeira in
the Pontus (near the modem Mksar, N. central IHirkey) in the
first century e.g., some time before the earliest in Greece or Italy.
There may be a simple explanation for this. The basic require-
ment for a water-wheel is a water supply which is steady all the
year round, and, if it is to be anything more than a toy, the quan-
tity of water needed is quite large. Mithridates' city was close to a
substantial river, the Lycus (modem Kelkit) which, though the
local rainfall is no greater than that of Greece or Italy, has a large
catchment area. Relatively few of the rivers and streams of Greece
and Italy (except in the north) maintain a substantial rate of flow
during the dry season. However, the effect of this geographical
fact on the history of the water-wheel should not be exaggerated.
Once the basic idea has been put into practice, the conservation
and management of limited or fluctuating water supplies follows
soon afterwards.

Our knowledge of Greek and Roman attempts to harness water
power rests on rather meagre evidence. Among the hterary sources,
Vitmvius (late first century B.a) is much the most important, and
he gives a dear description of an undershot wheel, which is dis-
cussed below. Two other alludons are important for the question
of dating. A Greek epigram in the Palatine Anthology (IX, 418)
speaks of the joyful release from drudgery which a water-mill has
brought to the women servants who previously had to grind by
hand. Its author was almost certainly Antipater of Thessalonika,
who was closely associated with a Roman noble family, the Pisones.
He lived and worked in Italy at the end of the first century B.C.,
and is probably referring to the installation of such a mill on a
country estate. His poem would be contemporary with Vitmvius'
work, but there is one interesting difference between the two.
Antipater speaks of the Nymphs (which personify the water) as
'leaping down onto the topmost part of the wheel'. Though this
has been disputed, there is really lit lie doubt that he is talking
about an overshot wheel — a more efficient type than Vitruvius'
— and this raises a question of priority, which will be discussed
later.

Glosely related to this is an allusion in Lucretius' poem On the

18 ENGINEERING IN THE ANCIENT WORLD

Nature of the Universe, where the poet is speaking about the move-
ment of the heavenly bodies (V, 509-^3, particularly 515-6). It is
a difficult and obscmie passage, but the gist is that one explanation
of the apparent diurnal rotation of the heavens is that a current
of air circulates around the umverse, causing the ^here' to rotate
*as we see rivers turning wheels and buckets {rotas atque haustrdf.
Since Lucretius uses this as an illustration, he clearly 2issumes that
water-wheels are familiar to his readers, cUid as he was writing
some 40 years earlier than Vitruvius and Antipater, this suggests
that the use of water power to work pumps (bucket-wheels or
bucket-chains, see Chapter 3) came earlier than its use for milling.

Other literary allusions add little or nothing to this. The arch-
aeological evidence is equally scarce, but very informative. Two
important wheel sites have been excavated — one in the Agora at
Athens, to the south of where the restored Stoa of Attalus now
stands, dating from mid or late fifth century a.d. The other is at
Barbegal, near Aries in southern France (just north of the Gam-
argue). A very big installation was built there by the Romans in

the late third or early fourth century, and was probably in use lot

the greater part of 100 years. It contained eight pairs of wheels,
each driving millstones in a mill-chamber beside the wheel-pit, and
its output would have been adequate not only for the 10,000 in-
habitants of Aries, but for some area around. The presence of a
Roman garrison might account for this. The remains are not very
extensive, but the main essentials can be reconstructed from them.
Evidence of an undershot wheel (in the form of chalk incrustation,
the wood having all disappeared) has been found at Venafrum in
central Italy, and a speculative reconstruction can be seen in the
technology section of the Naples Museum.

There are three basic types of water-wheel — the vertical-shaft,
the undershot and the overshot The vertical-shaft wheel has a
number of blades inclined at an angle of about 30^ to the vertical,
fixed to a hub near the bottom of the shaft. The water is directed
onto the blades through a wooden trough which slopes down at a
steep angle, so that the water strikes them at high speed. This re-
quires a situation where there is a drop of some 10-12 ft (3m) im-
mediately beside the water source. Sometimes a pit can be dug for
this purpose, but adequate arrangements have to be made for the
spent water to drain away from it. Since the shaft is vertical, it can
be made to turn millstones directiy, without any need for geara.

Copyrighted material

POWER AND ENERGY SOURCES

19

In the absence of any conclusive evidence^ some historians of
the subject have used the following argument : This is the 'most
primitive' form of water-wheel, so, since the Romans developed
the more sophisticated undershot and overshot wheels (for wMch

we have good evidence), we must assume that they started with
the vertical-shaft type. The parallel between this supposed
sequence and that attested for Renaissance Europe is also invoked
in support. This a priori argument is attractive, but it does rest on
two doubtful assumptions (a) that milling was the first operation
for which water power was used — and the passage from Lucretius
quoted above makes this very doubtful — and (b) that gearing of
some sort had not been previously invented for other purposes,
such as coupling animals to a water-pump. Archaeological evi-
dence (or rather, the lack of it) does not help to decide the question.
No certainly identifiable Greek or Roman remains of this type of
wheel have been found, but the entire structure, including all the
water-guidance system, would have been made of perishable
material. By contrast, the overshot wheel required a stone-built
wheel-pit, which has good chances of survival, and can be identified
as such.

The second basic type of wheel is the undershot, sometimes called
*Vitruvian' from that author's description (X, 5). It is highly
significant, and consistent with the evidence from Lucretius,
that he hrst introduces the water-wheel as a power source for
working a bucket-chain, and then says, 'It is sJso used for com
milling, the design being the same except that there is a gear-
v/hod on one end of the axle . . He makes no mention of the
vertical-shaft wheel. The structure he describes is very simple (Fig.

2) . It consists of a spoked wheel of unspecified diameter, with
vanes or paddles around its circumference (Vitruvius calls them
pinnae, a word used elsewhere to mean the wing-feathers of a

bird), which are driven round by the current in the river. There is

nothing in his words to suggest that a miil-leat was channelled off
for the purpose.

The third and most efficient type of wheel is the overshot (Fig.

3) . Using the same kind of argument as with the vertical-shaft
wheel, it is usually held that this was developed from the under*
shot wheel, the intermediate stage in this process being the so-called
lireast-shot' wheel, which is a simple modification of the under-
shot, the water bdng supplied through a trough level with the

20 ENGINEERING IN THE ANCIENT WORLD

Fig. 2. Undershot water-wheel.

Fig. 3. Overshot water-wheel.

axle, so that the main force on the paddles is from the water falling,
not merely flowing past. But there is another equally attractive
hypothesis — that the overshot wheel was conceived independently
of any other type, by amply revenuig the action of the bucket-
wheel. If one can put power Into that machine and get water out

Copyrighted material

POWER AND ENERGY SOURCES

21

of the top, why not put water into the top and get power out of it?
In fact, this possibility would be clearly demonstrated each time
somebody finished a spell of work treading a bucket-n^ed. It
would have to be slowly reversed until all its buckets were emptied,
and the pull it exerted during that operation would be clearly felt.
The bucket-wheel was certainly in use by Vitru\ ius' time, perhaps
for some while before, and though he docs not describe an over-
shot wheel, that might be due to the fact that the undershot type
was the only one he had seen.

The question of priority, then, is not easy to answer; but in
power ouq)ut and eihciency the overshot wheel is well ahead. The
structure required for an undershot wheel is simply a vertical wall
beside a river or stream,* and, if the water supply is limited, some
sort of partial dam to narrow the channel and make it flow more
rapidly in the region of the wheel. The potential power output
dq)end8 on two factors — the velocity of the water flow and the
area of the vanes on which the water impinges (the 'scanned area').

To take a simple example. If the area of each vane is l,000cm^
(just over 1 sq ft) we may assume that roughly this area is being
scanned at any one time. (The exact figure depends on the number
of vanes, the diameter of the wheel, and other factors, but this ^vill
do as a crude approximation.) If the water flows past at about 150
cm/sec (5ft/sec) the theoretical power available is about J h.p.
(186 watts), but as the undershot wheel can only be made about
22% efficient at the best, this would provide a real power of only
about ^ h.p., or half the power output of a man waking a tread-
mill. If the water flows twice as fast, the power is increased dg^t
tunes and things look better. The dieoredcal power available is
nearly 2 h.p., and the actual output might be about 0.4 h.p. — the
equivalent of four men. On the other hand, the water supply for
such a performance could not be obtained from anything less
than a small river, with a flow of (say) 125 gall/sec, which might

be around 12ft wide (3.5m) and 4in (10cm) average depth in
cross-section.

The overshot wheel can be made much more efficient — up to
65% or even 70%. Provided that the wheel revolves fast enough,
and the boxes are large enough to catch all the water as it comes
from the launder, most of the potential energy in the water can be

^For an example, see the Byzantine mosaic in the Palace of the
Emperors, illustrated in Antiquity XIII (1939) pp. 354-6 and Plate VII.

Copyriyhioa inaici lal

22 ENGINEERING IN THE ANCIENT WORLD

Utilized. This potential energy can be worked out quite simply
from the rate of the water flow, and the dq>th of fall which, as a
rough approximation, may be taken as equal to the diameter of
the wheel. The lesser rate of flow given in the last paragraph (31
gall or 140 //sec), if delivered to an oversliot wheel of 7 ft
(2.13 m) in diameter, would give a theoretical power output of
just under 4 h.p. (nearly 3000 watts), and an actual output of
perhaps 2-2^ h.p. The power of each of the sixteen wheels at
Barbegal might have been of this order. For a modern (and
slightly depressing) comparison, a small motor-cycle engine
(250 cc) develops about the same power.

Overshot and undershot wheels may usefully be compared in
two other respects — behaviour under extra load and cost of con-
struction. They behave in opposite ways under extra load. I^nce
an undershot wheel IS absorbing kmetic eneigy from die movmg
water, its torque depends on the difference between the velocity
of the water on arrival and the speed of the paddles. To put it
very simply, it generates power by slowing the water down. There-
fore, if extra loading is put on the wheel (e.g. by using bigger mill-
stones or putting bigger buckets on a chain) it will turn more
slowly, but will develop more torque. Conversely, the overshot
wheel has a minimum working speed, below which the water
begins to overflow the boxes and spill into the pit, reducing the
power output and efficiency. These factors must be taken into
accoimt when designing the wheel and the gearing, which will be
discussed later.

In pcnnt of cost, the undershot wheel has the great advantage
that no pit is needed, that any riverside atuation can be used (this

may reduce transport costs, which were high) and no engineering
is required to raise the water to the necessary height. This has to
be done for an overshot wheel, and may be very expensive. If the
gradient of the river bed is slight, it may be necessary to build an
artificial channel 200-300 yards long, support it above ground
level and make it waterproof. The cost of constructing the aque-
duct to feed the Barbegal system must have been very considerable.
The efiiciency of the undershot wheel is much less, but this need
not have worried the ancient engineers all that much. Where fuel
is e3q>ensive (as in the modem petrol engine) efficiency is the first
essendal, and must be achieved at almost any cost, but where the
energy source is running water, and *costs nothing*, the only

Copyriyhioa inaici lal

POWER AND ENERGY SOURCES

23

requiremeiit of a ^^led Is that it should deliver enough power to do
the ymsk. If the choice lay between an undershot wheel which
would just about turn the millstones and an overshot one which
would turn them faster, but would cost four or five times as much,

and might have to be built some miles away, the undershot would
be preferred. Where the water supply was too small for anything
less efficient than the overshot, there was no choice. This was
probably the case in the Atheni^ Agora.

Fig. 4. Water>mill gears with toothed wheds.

It is not difficult to see, from Vitruvius' clear description and
from the evidence of the Agora mill, how the water-wheel, turning
on a horizontal axle, was coupled to the upper millstone on a
vertical axle (Fig. 4.) 'A toothed disc identatiim) is keyed on
to the end of the axle, and turns in a vertical plane, at the same
speed as the wheel.' (The odd phrase in cuUrutn, which has not
been satisfactorily e3q>lained, is omitted from this translation.)
*GIo0e to this disc is another larger one» toothed in the same way
{item dentatum) and horizontafiy placed, with vfbldi it engages

Copyrighted material

24 ENGINEERING IN THE ANCIENT WORLD

{continetury Vitruvius is clearly talking about two crown wheels.
We use the word 'toothed' more loosely, of a flat cog-wheel with
radial teeth, but if one thinks of an animal skull, with curving
jawbone and teeth at right angles to the Vim% it be seen that
Vitruvius' usage is really mcnne exact. Marks made by the rim of
the vertical gear-wheel on the edge of the gear-pit in the Athenian
Agora mill confirm that the teeth were not radial. Vitruvius' ex-
pression 'toothed in the same way' {item dentatum) &u^^tst& that
the so-called lantern pinion (Fig. 5) with two discs was not

I

Fig. 5. Water null gears with toothed wheel and lantern pinion.

known to him; a Roman wood-and-metal pnnion of this type has
been found in Germany* but what part it played in mill machinery
(if any) has not been satisfactorily explainied. Doubt has been cast

on Vitruvius' statement that the horizontal gear-wheel (coupled

to the millstones) was larger than the vertical one on the wheel-
shaft, since this would mean that the millstone was geared down,
and turned more slowly than the water-wheel. Some scholars have
arbitrarily changed the text (from maius to ininus) to avoid this
problem. It is true that later European mills had the opposite
arrangement, the millstones being geared up by as much as 2^ : 1,
but these were Ing, powerful overshot wheels, and the type which

^Illustrated in L. A. Moritz, GrauMmUs and Flour m Qasskal AtOiqmty
(OUP 1958) Plate 14(c).

Copyrighted material

POWER AND ENERGY SOURCES

25

Vitruvius describes might not have developed enough power even
for a 1 : 1 gear ratio. The consequences are not nearly so disas-
trous as some historians suggest. The miller sunply worked more
slowly, and estunates of flour production should take this into

account.

Water-wheels were clearly used for water-raising and for milling.
We might expect some other applications, but the only evidence
we have is a brief and tantalizing allusion in Ausonius' poem on
the River Moselle, written about the middle of the fourth century
A.D. Speaking of the River Erubius (the Ruwar) he says :

'He, turning the millstones with rapid, whirling motion,
And drawing the screeching saws through smooth white stone.
Listens to an endless uproar from each of his banks' (362-4 )

Ausonius* style is not exactly straightforward, and it is difficult to
be sure exactly what he means, but he certamly seems to be saying
that water-wheels were used to drive saws for cutting stone. The

noise was incessant because the power-driv cn saws, unlike those in
an ordinary mason's yard, did not stop for a breather every few
minutes. PUny {Nat. Hist. 36, 159) mentions stone from this
area and from others which 'can be cut with a saw of the kind they
use for cutting wood — even more easily than wood, so they say*.
It was used for roof and gutter tiles, and was almost certainly some
form of soapstone.

But how did the river 'draw* the saws through stone? Ttahere
would be a strange word to use (even for Ausonius) of a circular
saw, though that was perhaps known in antiquity. TAd the wheel
have a cam and lever, or a crank and connecting-rod to push the
saw back and forth? In the absence of any evidence for either we
can only guess, and regret all the more that no technical writings
have survived from that area or from that period.

Mention was made at the beginning of this section of the de-
structive power of a river in spate. Though this power itself was
not put to useful purpose, some Roman mining installations in
Spain, by *imitating nature*, achieved a great saving of man-
power and time. Large reservoirs, known as 'hushing tanks*, were
constructed on the hillsides above the workings, with sluices at one
end which could be rapidly opened. When the tanks were filled (in
some cases via a fairly long aqueduct) the sluices were released»
and a great wave of water rushed over the workings, carrying away

Copyrighted material

26 ENGINEERING IN THE ANCIENT WORLD

with it large quantities of spoil. The same water supply, regulated
down to a steady trickle, could also be used for washing ores.

WIND POWER

Although the Greeks and Romans harnessed and used wind power
for sailing ships, they do not appear to have developed the rotary
windmill as a power source. This is strange, and no satisfactory

reason has yet been offered. They were perfectly well aware that
by adjusting the set of the sail a boat could be made to travel at an
angle to the direction of the wind, and a very slight development
of this idea could have led to the type of sail-mill to be seen nowa-
days on Mykonos and in Crete. But we have no evidence for any
such machine in classical antiquity. The one and only mention <^
harnessing windpower is in the Pneumatica of Hero of Alexan-
dria (1, 43), and, being unparalleled, it has come under suspicion
as a later interpolation. But there is nothing in the vocabularly or
style of the Greek which is inconsistent with the rest of Hero's
works, nor is it easy to see what motive could have prompted any-
one to insert such a passage Into a fairly well-known text some
time in the Middle Ages, when the windmill had come into
general use.

Hero's machine, in which wind power is used to blow an organ,
is crude but workable. Only a very sketchy outline of the instru-
ment itself is given, with no mention of a keyboard or air reservoir.
This may mean that it was something like an * Aeolian harp' (the
introductory sentence says *it makes a noise like a pipe when the
wind blows') or perhaps we are meant to fill in the details from
the very full description of an organ given in the previous chapter.
The air piunp consists of a piston and cylinder, the piston compress-
ing on the down-stroke (Fig. 6). No valves are mentioned, but we
must assume the same kind of arrangement as that given in
Chapter 3, except that the cylinder is inverted. It is worked by a
rocker-arm, which has on its opposite end a small horizontal plate.
The windmill itself is mounted on a separate base, so that it can
be turned round as required to face the wind — perhaps through
an arc of 90°. It has a single axle, with two discs {tympania —
'little ch ums') on it. One has projecting radial rods which, as it
turns round (anti-clockwise in the diagr<im) push down the small
plate and lift the piston. As each rod slips off the plate, the piston
is allowed to fall, and its weight then forces the air out into the

Copyrighted material

POWER AND ENERGY SOURCES

27

organ ppes. The other disc on the same axle is fitted with *vanes,
like the so-called anemouria\ The word translated as *vanes*
iplatax) is used elsewhere to mean oar-blades, which suggests that
they were wooden, and rigid. The word anemourion docs not

occur anywhere else except as a proper name for a promontory in
Asia Minor, but must mean something like * wind-fan'.

Fig. 6. Hero*s windmill for blowing an organ.

Here we have a crude and rather inefficient substitute for the
cam, converdng the rotary motion of the windmiU to the up-to-
down motion of the pump. We are not told how many radial rods
were used — probably two at the most, since the interval between
each thrust would otherwise be too short to allow the piston to
empty the cylinder. Hero says 'they (the rods) strike the plate at
longish intervals' (ek dialeimmatos), which also suggests that the
windmill was designed to turn slowly, with a pitch of perhaps only
5-10° on the vanes.

The device was clearly a toy, but why did nobody (apparently)
see its potential as a power source ? Perhaps its scale was the real
reason. A power source, by definition, had to be something which
could i«place a man or a small animal — that is, something which
developed about ^ h.p. at least. It may well be that a small wind-
mill, with rigid wooden vanes, was simply not thought of in this
category, and nobody tried experimenting with a bigger and
better one. But it is still very puzzling.

Copyriyi uuj I . id^Li lai

28 ENGINEERING IN THE ANCIENT WORLD

STEAM POWER

The failure of the Greeks and Romans to harness steam as a power
source was without doubt one of the many factors which prevented
industriaHzation in thdr society. How near they came to developing
a workable steam engine is a much-debated question.

Once agam it is Hero of Alexandria who provides the only
mention in the literary sources of devices worked by steam {Pneu-
matica II, 6 and 1 1). The second of these is 'a ball which spins
round on a pivot when a cauldron is boiled*. He does not give

this device a name, though aeolipyle (or aeolipile) i.s suinetiniei)
used — mistakenly, because that was a different device ahogether.
The design is simple (Fig. 7). Pressure builds up in the cauldron,
and steam passes through the pipe FGH into the sphere, from
which it escapes at various points, but mainly through the bent
tubes I JK and LMN. As the steam is forced out in one direction
(from the outlets), it causes a reaction thrust in the opposite
direction, and makes the sphere revolve. The principle is that

Copyrighted material

POWER AND ENERGY SOURCES 29

of jet propiilsioiiy and the device described in Chapter 3 of the
same book, in vrbkh. figurines are made to revolve inside a trans-
parent altar, worlcs in the same way, except that the expansion of
heated air is used instead of steam.

Could this form of steam engine ever have been used as a prac-
tical power source ? The answer is, almost certainly not. It operates
best at a high speed, and would have to be geared down in a high
ratio. Hero could have managed that, since the worm gear was
familiar to him, but not without friction loss. Inadequate heat
transfer from the burning fuel to the cauldron would keep the
efficiency low, but the worst problem of all is the 'sleeve joint',
where the pipe FGH enters the sphere. When making a workmg
reconstruction oi this device, I had the greatest difficulty in reach-
ing a compromise between a loose joint which leaks steam and
lowers the pressure, and a tight one which wastes energy in friction.
It is in the realm of possibility that, given the technology of Hero's
age, overall efficiency might have been as low as 1 % . If so, then
even if a large-scale model could have been built, to deliver yq
h.p. and do the work of one man, its fuel consumption would have
been enormous, about 25,000 B.T.U. (26.8 X 10® joules) per hour.
The labour required to procure and transport the fuel, stoke
the fire and maintain the apparatus would have been much more
cxpexisiwe than that of the one man it might replace, and the
machine would be much less versatile.

In hb introductory chapter, Hero speaks oi his various devices
as providing *some oi them useful everyday applications, others
quite remarkable effects'. We must conclude that the steam engine
came into the second category. Its most remarkable feature is in
fact the speed of rotation. My own working model has achieved
speeds of the order of 1,500 rpm, and, with the possible exception
of a spinning top, the ball on Hero's machine may well have been
the most rapidly rotating object in the world of his time.

It is true that this toy (as it may justly be called) does not incor-
porate the essential elements of a useful steam engine, but it is
equally true that all those elements are to be found in various
otiier devices which Hero describes. To make a conventional steam
engine it is necessary to develop techniques of making metal
cylinders, and pistons to fit them; but this problem was tackled in
the design of the force pump, and there is even a posability that
'lapping' was used (see p. 76). It is not In fact necessary to have an

Copyrighted

30 ENGINEERING IN THE ANCIENT WORLD

efficient method of convertiiig rectifiiiear to rotative motion for the
construction of a basic steam engine. The earliest working ones
were the beam type, which worked piston pumps without cranks
or rotative motion. The one other essential is the valve mechanism,
and Hero had devised one for what is usually known as 'Hero's
Fountain' — a device exactly like a modem insecticide sprayer, in
which Kquid is forced out of a container by compressed air {Pneu-
matica I, 10.) Water had to be controlled under pressures of the
order of 5-6 Ib/sq in (0.35 kg/cm^). The type of valve he used is
not ideal for steam, but it would have done to start with, and
might have been improved in the light of experience. An even
more agnificant feature of the 'fountain' is that Hero uses a
spherical pressure-vessel on a special stand, showing clearly that he
was aware that sheet metal of a given thickness will stand up to
greater internal pressure in that form than in any other. In the
progress of boiler design, this might have been the first advance
on ^e bunged-up cauldron.

But there was no progress — not even a beginning. What Hero
failed to do, and nobody else apparently tried to do, was to com-
bine these essential elements — boiler, valves, pistons and cylinder
— to form a steam engine. Why did he not? Perhaps he never
thought of reversing the action of a piston pump, forcing liquid or
air into the cylinder and taking thrust from the piston. One toy
which came very near to this was a 'jumping ball' {Pneumatica
II, 6), in which a light ball (made of thin sheet metal ?) was blown
up into the air by a jet of steam from a boiling cauldron. But where
eacpanding hot air, or compressed air, was used to move something
mechanically, it was done eidier by inflating a bladder, which
lifted up a wdght, or else by shifting water or mercury from one
side to the other a counterbalanced system, which then swung
up or down and operated the mechanism by chains and pulleys.
TTiese two methods are exemplified in Pneumatica I, 38 and 39.

Another explanation which has been suggested is that Hero did
attempt to combine the elements of a steam engine, and either
blew himself up in the process, or was frightened off the idea. But
if the first of these occurred, it is strange that there is no mention
of it in ancient tradition. It would have been such a good caution-
ary tale for anti-materialist philosophers. The second does not ring
true psychologically. Stephenson, Diesel and Whittle each per-
sisted with his engine despite nasty accidents and nairow peisonal

Copyrighted material

POWER AND ENERGY SOURCES 31

escapes. Hero was nothing if not an enthusiastic inventor, and is
likely to have thought, as they did, that 'it would never happen to
him*.

The other technical reason offered for the failure of this develop-
ment (as opposed to economic and social factors, which were
probably the real causes) was the lack of high-quality fuel. This
has implications in a wide range of other contexts, particularly
metallurgy. The two fuels in general use were, as one would natu-
rally expect, wood and charcoal (which was called anthrax in
Greek and carbo in Latin). Charcoal was preferred for cooking,
because it burned more slowly and made less smoke than wood,
and, since artificial draught was used (a fan or bellows), the tem-
perature would be controlled more easily. It was the nearest
ancient approach to 'the Heat that Obeys You'. It was also used
in metal smelting furnaces. Being almost pure carbon, it is capable
under ideal conditions of reaching a very high temperature and,
unlike wood, does not contain cellulose or water, both of which
slow down the oxidization and limit the rate of heat production.

It is widely held that the Greeks and Romans were unable to get
their furnaces much above 1150°C, and that this hindered the
development of their iron technology, but recent experiments have
suggested that this view is quite mistaken. In a furnace modelled
on a Roman type temperatures in the region of 1300°G were
reached. If inexperienced modem operators could do this at the
first attempt, the skill derived from a year or two of trial and
error could surely have produced better results. Such a furnace
represents an ^equilibrium situation*, in which the temperature
rises until the point is reached at which the total heat losses exacdy
balance the amount of heat being generated. So, to raise the tem-
perature, it is necessary either to increase the heat production, or
cut down the heat loss, or both. As far as the first is concerned,
little could have been done in the ancient world without the
chemical knowledge by which to improve the fuel. On the other
hand, charcoal has a calorific value of about 12,000 B.T.U./lb
(27.8 X 10® joules/kg) and compares favourably with coal. What
they could and did do was to cut the heat losses, by designing an
effective furnace, by management of the air draught and, above
all, by the charging technique — selecting the right proportions
of ore and fuel, stacking them in the best way, and replenishing
the fuel at the later stages without setting up 'cold spots'. None of

Copyriyi uuj I . id^Li lai

32 ENGINEERING IN THE ANCIENT WORLD

this really requires scientific luiowledge» merely patience, careful
observation and a lot of practice.

In point of cost, charcoal comes out quite wdl as a fuel. It is
dearer than ivood, but by how much we cannot really say. Since
it is produced by the slow partial combustion of wood, without the
addition of any other substances, the diflPerence would be largely
accounted for by the labour costs, arid it is notoriously dilTicult to
assess the impact of such costs in the ancient world. Not all kinds
of wood are suitable. The best are the hardest and most close-
grained varieties, such as holm-oak (ilex) and beech. This might
add transport costs, since charcoal-burning could only be done in
the long term in areas well supplied with such trees, which might
be some distance from the main market.

Goal was used as a domestic heating fuel in some parts of the
Roman em|»re (particularly in Britain) but it never made more
than a marginal contribution to fuel resources. There is no evi-
dence for deep mining; all the coal used was outcrop, and probably
of rather poor quality. It was not normally used m smeking f iu>
naces, though Pliny (Nat. Hist. 34, 8, 96) does apparently mention
its use in copper casting, which can be done at a considerably lower
temperature than iron smelting. In fact, it was not until the inven-
tion of coking in the seventeenth century that coal really super-
seded charcoal as a smelting fuel, and coke stands to coal in much
the same relationship as charcoal does to wood — it is much less
dense, and has a porous structure which exposes a large area to the
air and makes more rapid burning possible.

Charcoal-making was, therefore, a very important activity in
the ancient world, but we have very little evidence indeed of the
methods used or the degree of competence readied. It was a craft
industry, like so many others, and it is most unlikely that any
technical manuals were ever written. We do have, however, in a
comedy of Aristophanes, a brief glimpse into what might be called
(rather pompously) the sociological aspect.

Charcoal-burners in the ancient world, like their few remaining
successors today, were normally self-employed. They tended to be
individualists of sturdy independence and, because they lived and
worked in wooded areas of the countryside, unsophisticated and
rather cut off from the current trends of city life. This is partly why
Aristophanes chose them to form the chorus of his comedy, The
Aeharnians, named from what was then a small village in 'char^

Copyrighted material

POWER AND ENERGY SOURCES 33

coal country* about 1 1 km north of Athens. One rather misleading
element in their characterization is that they are all very aged,
since they claim to have distinguished themselves at the bat^k of
Marathon some 66 years before, which would make them at least

85 or so. But Aristophanes had other very good reasons for making
them elderly, which have nothing to do with their occupation,
and we should certainly not infer that it was a dying rural craft,
confined to the older generation. They are violently nationalist —
meaning, in the context, fanatically anti-Spartan — and they
attack the envoy who has been sent to negotiate a secret peace
treaty. After an unpleasant brush with them (off-stage) the envoy
calls them 'Senior citizens of Achamae, compressed old blocks,
ilex-maple-hardwood-veterans of Marathon'. The second of these
abusive terms, by the way, probably refers to a process of treading
charcoal to make it more dense. When they make their entrance
soon afterwards, they live up to this image, since they violently
attack the hero, who has dared to make peace terms with The
Enemy. He brings about a temporary lull by parodying a scene
from tragedy which would have been familiar to the audience ; he
threatens to 'slay with the sword a hostage he has in his house'.
Who can this be ? One of their children ? It turns out to be a char-
coal-burner's pot {larkos), for which they express a deep senti-
mental affection. 'Spare him T they cry, *he is one of us, from our
own village T So the hero escapes, but it was a near thing. Later^
when an is explained and forgiven, the chorus smg a little song,
ostensibly in their own character, but really in the poet's :

Come hither Muse, Ac ham i an Muse,

Whose song has the vigour of burning fire

M a king the sparks fly aloft from kolmroak charcoal

By the fanned breeze agitated.

And the Uttie fish are laid out for the grUUng,

And cooks stir up the Tartar sauce

A gleaming garland for the forehead of a lovely sprat.

And others knead the barley-cake;

Come with a rousing rustic song.

Come to me, O Muse — you are one of us!

(665-675)

Copyrightcu \ ,,ao^t lal

2

Water supplies and engineering

Water supply represented one of the most serious problems for
Greek and Roman urban communities. With the foundation of
each new centre, and each major expansion of an existing one, new
supplies had to be found, tested and conveyed — perhaps over some
distance — to the delivery point, and at least some storage capacity
had to be provided. This might require extensive building works,
which had to be planned, inspected and maintained, and inevi-
tably all kinds of l^;al and administrative problems arose, over
both the legitimate rights of consumers, and the illegal activities
of mains-tappers and corrupt officials.

It is not surprising, theref<»«, that \^truviu8 devotes a whole

book (the eighth) of his De Architectura to the subject, and, since
this represents a compendium of the theoretical knowledge and
practical experience of his day (first century B.C.), it affords a con-
venient framework for this account. It must be remembered, how-
ever, that Vitruvius and Frontinus (the other very valuable source)
are writing — from personal knowledge — about the water supply
to Rome in the last century B.C. and the following century, which
was by far the biggest and most complicated in the andent world.
Some of its features, therefore, may have been unique — for
example, the division into three types of supply, and the vecHgal
or water-tax levied on private consumers.

The theoretical basis of hydrostatics, which \^truvius draws
mainly firom Greek scientific writers such as Gtcnbius and Archi-
medes, is discussed in Chapter 8. The meteorology (such as it is)
comes from even earlier Greek writers. For instance, the account
in his second chapter of water evaporation and precipitation differs
very little from the kind of popular science expounded by the
Sophists in the late fifth century b.c., and parodied by Aristo-
phanes in The Clouds. But apart from this general background,
there arc a few specific principles of physics (or physical chemistry)
which are derived from Greek medical writers. One deserves

Copyriyhioa inaiciial

WATER SUPPLIES AND ENGINEERING 35

special mention here, and the rest nvill be disciissed as tiiey arise in
the course of the account
At the end of Chapter 1 (and elsewhere) Vitruvius asserts that

the influence of sunlight on water is detrimental, since it causes the
purest particles to evaporate and scatter, leaving behind the 'heavy,
coarse and unhealthy parts'. This is of course perfectly true, and
easily observable. One of the tests for purity listed in his fourth
chapter — -boiling a sample of the water in a metal pot, and seeing
whether solid impurities come out of solution — is simply an inten-
sified version of the same process. He does point out, however, that
the polluting effect of the sun is most obvious when the water is
eaqposed cm level ground, and ibmiort static. It then causes the
growth of algae and insect larvae. The andents, having no micro-
scopes by which they could see the eggs, believed that such crea-
tures were generated spontaneously by the action of heat on mud.
The warmth also causes the production of ^poisonous vapours^
• — methane and other gases given off by rotting vegetation. All
these phenomena (familiar nowadays to amateur water-garden
enthusiasts) combine to make water from marshy sites highly un-
suitable for drinking, in the unanimous opinion of all ancient
authorities.

A curious converse to the 'heat-pollution' idea was expounded
by the Greek physician Hippocrates (fifth century b.c.) in his
treatise on Ahs, Waters and Places — namely, that freezing like-
wise pollutes water, by removing the purest particles, leaving the
impurities behind. Though the experiment which he adduces in
support (freezing a quantity of water in a metal jar, then thawing
it and measuring the loss of volume) is crude and misleading, it is
not difficult to see how the idea might have grown up. However
clear and pure water is, it becomes opaque when frozen. In the
ancient world it must have been very difficult to collect ice or
snow (which they used for cooling drinks) witliout picking up some
dust or grit with it. F.ven today, however carefully one cleans the
refrigerator drawer, and however clear the water one puts into it,
there always seems to be a spot of sediment at the bottom of the
gin-and-tonic when the ice has melted. Though Vitruvius does on
occasion speak of water supplies originating from melted snow or
ice, he does not mention the 'pollution effect*.

His account of water supplies begins at the basic and practical
level, with some procedures for locating underground sources of

36 ENGINEERING IN THE ANCIENT WORLD

water. The first and simplest method is to search for water vapour
rising from the ground. This is best done at sunrise, ^en the mdst-
ure has risen to the surface (by capillary attraction) during the
night, and evaporates as soon as the scnl surface is wanned. The

best way to observe it, says Vitruvius, is to he face down on the
ground and look along the surface, where the refraction (lie calls
it 'moisture forming curls and rising into the air') can be most
easily seen. This is a sure sign of the presence of water, and justifies
a test dig in the area.

Vitruvius then digresses on the types of soil in which water can
be found. In ascending order of merit they are (1) day (limited
quantity, bad flavour), (2) loose gravel (water at greater depth, and
muddy), (3) peat (water gathers in small droplets, and only collects
if there is a sealing basin beneath), (4) gravelly soil and sharp sand,
(5) red sandstone rock, provided that the fissures do not drain too
much away, and (6) at the foot of a hill, among hard rocks — the
best type of source. The only suitable one to be found on an ex-
panse of level ground is an artesian well, with its outlet in a shaded
position.

Another good indicator of water resources Is plant-life. Vitru-
vius lists six in particular which will only grow in consistently
damp scnl — bullrush, wild osier, alder, withy, reeds and ivy. Oi
course, they usually grow in marshy sites, which are unsuitable as

water sources, but where they grow in other situations, and in the
suitable soils, they mark a possible source. Needless to say, they
must have grown there spontaneously, and must not have been
transplanted.

Digging a well is a long and laborious buaness, and to dig one
which turns out to be useless is extremely frustrating. So Vitriivius
suggests some additional tests to be carried out before finally start-
ing on the actual digging. If the indications already mentioned
are present, a pit should be dug about 1 m square and 1.5 m deq>.
A metal basin is to be placed in it, upside down and smeared with
oBvc oil on the inside. The pit should then be covered over wth
reeds or tree-branches with the leaves still on, and a light covering
of soil on top. This should be done in the evening, and the covering
left undisturbed overnight. In the early morning it should be
opened up and the basin should be examined for droplets of con-
densation — the olive oil would make these more easily visible. If

Copyrighted material

WATER SUPPLIES AND ENGINEERING 37

there is a dear indication of water vapour, it is almost certainly
worth Yfialt to sink a well«fihaf t Other objects may be put into
the jnt to detect condensation. A unfired clay pot, which is
hygroscopic, will go soft (and possiMy collapse) during the test A

woollen fleece will not attract so much moisture, but even a small
quantity can be detected by wringing it out. Finally, an oil lamp,
filled with oil and lit the night before, will not burn completely
dry.

Obviously, the ideal source of water is a spring on a hillside,
where the water flows continuously through a natural outlet. It
can be located without speculative digging, and the height makes
it possible to convey the water over some distance by gravity-flow
systems. Such sources are not uncommon in Greece and Italy,
where limestone formatimis afford cavities for underground
storage. Obvious examples are the Appenine foothills in Italy, and
in Greece the range between Attica and Boeotia extending west-
wards iii far as Parnassus, and also the hills of the Argolid and
southern Peloponnese. On many of these hills the average rainfall
is not much less than that in the drier parts of Britain or many
eastern states of the U.S.A., but it is mainly concentrated in the
winter months (October-March), which makes storage capacity
rather more important than catchment as a limiting factor on
supply. The writer of the First Delphic Hymn calls the Gastalian
spring 'much-watery'. This may sound like a statement d the
obvious, but it goes on bubbling up in August after months of
drought, and it is easy to see why he took the trouble to mention it.

There are two forms of conduit in which water can be ccmveyed
by gravity flow. The open conduit — much the more common in
Roman systems — consists of a channel, usually built into a stone
structure and waterproofed with plaster or cement. To keep the
level of water even, it has to slope at a more or less consistent angle
along its entire length. The gradient was normally between 1 in
150 and 1 in 500 (Vitruvius recommends 'not less than 1 in 200').
A closed conduit usually took the form of a round waterproofed
pipe of metal or earthenware, completely filled throughout with
water. It can slope up or down at any angle, provided that it does
not rise at any point above the level of the intake. Ancient systems
on the whole tended to be either all open or all dosed, but occadon-
ally a combination ol the two forms was used.

The basic problem ai the open-channel system was that the

38 ENGINEERING IN THE ANCIENT WORLD

gradient had to be maintained consistently over the rises and falls
of ground level between the source and the delivery point. If a hill
intervened, there were two ways of coping. If there were suffident
masons available, and a copious supply of local stone, a channel
was built around the hillside, followii^ the contour line apart
from the slight fall required for the flow. It would be supported on
what was, in effect, a broad, low wall, with faced stone on the
outside and a rubble in-fiU, with thin slabs of stone forming the
bed and sides of the channel, and a lining of cement to make it
waterproof. The Romans called this a substructio (Fig. 8).

Coverkig
sUb\

Water channel
lined with
nwrtar

'Brick, or
^asMar stone
faclnff

Concrete or

Fig. 8. SiAsimetio in Roman aqueduct

It had three very serious drawbacks. First, it could be very
extravagant of materials and labour, since it would have to be
built on the bedrock or a very firm subsoil, to avoid the danger of
sections being carried down the hillside by heavy rains or land-
slides. It was very much exposed to pollution, even if covered over
with stone slabs ^ — ^ another expensive measure. And finally — a
vital consideration for many Greek and Roman cities — in the
event of a siege, it was extremely vulnerable to enemy attack. The
alternative of tunnelling through the hill was therefore generally
preferred.

The usual scheme, as recommended by Vitruvius, was to make
the tunnel more or less straight, with vertical shafts up to the sur-
face at intervals of about 116ft (35.5m). This would seem to be a
rather excessive provision of shafts, and a number of e3q)lanations

Copyrighted material

WATER SUPPLIES AND ENGINEERING 39

have been put forward for it. One is that the work-force used on
such projects — or at least their supervisors — mig^t have had
mining experience, and since the usual ancient mining technique
was to sink shafts and join them together with horizontal tunnels

('adits'), this might be a simple transference of the technique from
one type of project to another. But the sinking of shafts can be
amply justified on two grounds — one applying at the surveying
stage, and the other after completion.

It is quite easy to ensure that a shaft is exactly vertical, simply
by hanging a plumb-line from a rod across the top, and seeing
that the bob hangs in the centre all the way down. If, therefore, a
line of posts can be laid (by optical sighting) up over the hill, and
shafts sunk from them, the horizontal alignment of the tunnel
becomes much eaaer. Once it has met up with the first shaft, it
can be aligned by sighting rods under the centre of each shaft, and
will more or less reliably meet up with the next along a straight
line. There is evidence to suggest that they did not trouble to get
the gradient exactly right at tlic initial stage, but corrected it later
by making a channel in the floor of the tunnel, which could be
adjusted a little way up or down as required.

Once the tunnel is made, the air shafts afford easy access to any
part of it for inspection and maintenance. An experienced miner
could, by r^;ular examination, spot the points at which subsidence
or collapse mi^t be expected, and any leakage from the water
channel could be promptly stopped. If a major fall took place and
a portion of the tunnel became flooded, it would be much easier
and safer to locate the exact point of collapse and the extent of
the flooding by lowering an observer down a shaft than it would
be to send him along the tunnel from the lower end. Finally, the
shafts would serve to release pockets of air-pressure which might
possibly form if the inflow of water increased very sharply (for
instance, after a freak rain-storm) and filled the whole tunnel.
Though the shafts might have been expensive in terms of the
labour required, they need not have affected the time-schedule
very much. The simple fact which makes any tunnelling a stow
job is that only a very few men can work on the face at any
one time. If all or most of the shafts were dug simultaneously, the
wodL-force could be more efliciently deployed for more ik the
time.

This type of water-tunnd had been familiar for some time in

Copyrighted material

40 ENGINEERING IN THE ANCIENT WORLD

the Near East before it was first used by the Greeks, and in many
modem textbooks (perhaps in recogmtion of this fact) it is called
fay the Arabic name qanat. The Greeks called it simply an 'excava-
tion* {orygma)f and the Romans called the tunnel a *cave' {specus)
and the shafts Svells* (putei). (Fig. 9.)

One of the most famous Greek examples was built in Samos in
the third quarter of the sixth century B.G., when that island was
under the rule of the tyrant Polycrates. According to Herodotus
(III, 60), the architect was called Eupalinos, and came from
Megara, where a similar but smaller tunnel had been built some
time before. Substantial remains of Eupalinos' project can still
be seen. The tunnel passes through a hill which is almost 1000ft
(300m) high. The length is about 1200yds (1100m), and it is
roughly square in cross-section, about 8 ft (2.5 m) in height and
width. The point at which the tunneUers» starting from each end,
eventually met can be seen, and the smallness of the error in align-
ment shows how competent the surveying mtist have been. Re-
mains of other tunnels are few and far between, for obvious reasons.
If the channel is not properly maintained and checked, the water
over a period of years erodes the walb of the tunnel, li a serious
collapse occurs, the upper part of the tunnd becomes flooded, and
the water eventually undermines the whole structure. All that
remains in many cases is the tell-tale line of holes at regular inter-
vals in the ground above.

Where the ground falls away below the level required to main-
tain the gradient, the channel must obviously be supported above
it. Up to a certain height — around 6^it (2m) a substruction as
described above, was used. At greater heights it becomes more
economical to use arches to support the channel, and the very
familiar pattern of the aqueduct (in Latin, arcuatio) takes over.

Arcuaiio

Fig. 9. Roman aqueduct.

WATER SUPPLIES AND ENGINEERING 41

It should be noted, however, that most of the extended water-
channels ran for much the greatest part of their length under-
ground. A typical example is the Aqua Claudia at Rome. Of its
total length of more than 35 miles (57km) only about the last one-
seventh was raised on arches.

The construction was usually very simple and straightforward.
Either bricks or cut squared stones were used to form the casing
of the pillars, with a rubble or concrete filling in the middle.
Where the aqueduct passed over a river, the pillars standing in the
river bed, and those on adjacent low land if it was subject to flood-
ing, were built with wedge-fihaped projections (*cut-waters') to
break up the force of the current Ilie structure containing the
channel, which ran above the arches, was usually of the same
materials. There was, however, a limit on the height of pillars
constructed in this way. It was possible for a very tall one (to put
it in die simplest terms) to fold sideways in the middle. This could
happen in a very high wind, or if subsidence took place at the base,
and if one pillar gave way, it could cau^e a progressive collapse of
the whole series of arches.

The Roman solution to this problem was to limit the height of
the arches to about 70ft (21m) or thereabouts, and when they
were working near to this limit, they made tlic pillars very massive,
and the arches between them narrow. If a greater elevation was
needed, they built the arches in two tiers, the jnllars of the upper
resting direcdy on those of the lower. The arches of the lower tier
could be made simple and not very heavy, their sole purpose being
to brace the pillars from each side. They conssted of the solid,
wedge-shaped stones (Voussoirs*) forming the arch itself, with
shaped stones forming a level top course above the arch. The
structure above the upper tier was exactly like that on a single-
tier aqueduct. A very good example of the double-tier type survives
at Segovia in central Spain, and is still used for part of the town's
water supply — a considerable testimony to the skill of the de-
signers and builders. It rises to a maximum height of 164ft (50m}
above the groimd.

When the aqueduct had to cross a very deep valley, the same
principle was carried a stage further. By far the most famous sur-
viving one — the Pont du Gard near Nhnes in southern France —
has two tiers of arches, with an additional structure of much
smaller arches on top, and the total hdght above the river bed is

Cc, , y od material

42 ENGINEERING IN THE ANCIENT WORLD

180ft (54.8in). The highest tier is made in the same material as the
rest, and carries a water channel about 4}ft (1.36m) wide and
5}ft (1.66 m) deep. During the many centuries when water flowed
through it» a thick incrustation of calcium carbonate has been
deposited on the sides and bottom. Even at this great height, stone
slabs were hoisted up and placed over the channel, to shield it
from the sun and from pollution.

Ancient water-engineers, especially the Romans, have been
subjected to criticism which, being ill-informed, is predictably
bitter. It is alleged that they built these elegant and massive
structures across valleys unnecessarily, having failed to realize that
*water finds its own level', and that a pipe could have been tciken
(for instance) down into the valley of the Gard and up the other
side. This criticism is wide of the mark on two points. First, it is
absolutely dear from the writings of Archimedes, Hero and A^tni-
vius that they all fully understood the pressure^uifibrium prin-
ciple, and second, the closed-pipe system is in many ways a much
less satisfactory answer to the engineering problems. It is less
expensive, but very much more difficult to construct, requiring
very specialized skills. It is unreliable, and subject to frequent
bursts and leakage. Once constructed, the conduit itself is not
accessible for maintenance, and if it becomes blocked, it may have
to be completely dismantled and rebuilt. An open-channel aque-
duct, by contrast, can be inspected and cleaned regularly, and, as
Frontinus points out (II, 124), it is even possible to rig up a tem-
porary by-pass, and carry out repairs on a short section without
turning off the main supply.

Two materials are suggested by A^tnivius for the pipes of a
closed system — lead {plumbum) and earthenware. He prefers the
latter, for several reasons. First, there is a danger of l«ui pdson-
ing from the formation in lead pipes of white lead oxide (which he
calls ccrussa) — a highly toxic substance- — and as corroborative
evidence he points out the unhealthy symptoms shown by workers
in lead-smelting and casting. Second, it requires workmen with
specialist skills (still called 'plumbers', though they now work in
copper and plastics) to carry out construction and maintenance,
whereas an ordinary bricklayer can deal with earthenware pipes
— or so he says. Third, but not least, lead is a much more expensive
material. Nonetheless, he gives detailed instructions for the making
of lead pipes.

Copyrighted material

WATER SUPPLIES AND ENGINEERING 43

The Rotuui method (which can be seen in the remains at Bath
in Somerset and in many other sites) was to use a rectangular sheet
of lead, which was folded (presumably around a wooden former)

into cither a circle or a triangle with rounded corners. The two

edges either had a simple overlap, or were overlapped and folded
(Fig. 10). For special soldering jobs, Hero of Alexandria recom-

fotdtd and

Fig. 10. Roman lead pipes.

mended pure tin, and lead/tin alloys were also used, but if the
join along a pipe was made simply by melting the lead or dripping
molten lead on to the join, it was probably not very strong or en-
tirely waterproof. The pipes were made in lengths of ten Roman
feet, i.e. 9ft 8^ in, or 2.95 m. There were 10 standard sizes, each
named from the width of the ^eet of lead used — that is, the cir-
cumference plus the overlap, not the diameter as specified now-
adays. The sizes are measured in 'digits', this unit being i^th of a
Roman foot — 0.73 in or 1.85cm. From the weight of lead which
Vitruvius specifies for one length of pipe, it can be seen that
(according to him) the lead sheet ought to be cast, or cast and
rolled, to a standard thickness of just under \'m (0.247in or
6.27 mm) regardless of the pipe diameter, which is a little sur-
pming. The sizes in modem units are given in Table 1. The lengths
were joined together either by butting them end-to-end and solder-
ing a collar around, or by flaring one end and tapering down the
other, inserting the taper into the flare, and sweating the joint
together by the application of heat, but how this was done is not
clear.

Earthenware pipes {tubuli fictiUs) were made in shorter lengths,

Copyriyhioa inaici lal

44 ENGINEERING IN THE ANCIENT WORLD

TABLE OF LEAD PIPE SIZES

Koman

Lead required per
^iO'fiHjC length

Diameter, aUming
for overlap •

lb

Kg

in

CItl

inn — ri inlt

zz-e

G9i

SiS'J

17'9

tO'9

27-^

3^6

tsj

22

2S9

e*3

iB

IJZ

ys*s

toz

130

S9

\

10 "digit

8G

39

! ^7

4-3

S -'digit

72

32*7

3

S "digit

} o-a

1*32

\ Overlap doubtful, so
\ very approximate In
\ thisnmge

A zo-digit pipe required i ton per its ft appro^a,

I tonnet per 3j'S m

TaUei.

but with much thicker walls — Vitruvius reccxmnends ^ot less
than two digits' ( 1 .46 in or 3.7cm). Each section had to be *tongued'
— that is, drawn in to a smaller diameter at one end. This was
probably done on a potter's wheel, in which case the length of
each section would be limited to about 3-4 ft (1-1. 2 m). To seal
the joints, Vitruvius suggests 'quicklime worked up with olive oil'.
He concludes with a characteristically Roman tip — crude but
practical. When the pipeline is complete, and the water is first
let into it, some wood ash should be thrown into the tank at the
supply end. This will find its way into any cracks or leaks in the
system, and help to clog them up ('grouting' is the technical term).
Some modem preparations for curing leaks In car radiatois work
on exacdy the same principle. In the Pagamon water system (to

Copyrighted material

WATER SUPPLIES AND ENGINEERING 45

be discussed shortly) some or aO of the joints between pipe sections
were enclosed in rectangular stone blocks, the pipes passing
through a round hole just above the middle of each block.

The two problems with closed-pipe systems are pressure and
sediment. If the pipe at any point falls a long way below either the
source or the delivery point, the water develops a pressure which
works out at approximately 4.41b/sqin for every 10ft of head, or
just under 1 kg/cm^ for every 10m head. If this pressure rises abofve
the order of 501b/sqin (3.5kg/cm') it to have several un-
pleasant effects. In lead pipes it tends to split open the join, and it
will soon find out any flaw or weakness in earthenware pipes. Also,
with both types it will tend to blow apart die joints between
sections. This is not such a serious problem when they are all in a
straight line, or curve giadually up or down, since the joints are
held together by the weight of the system as a whole. But Vitruvius
points out that if there is a sharp bend between a vertical or near-
vertical section and a horizontal one, there is a great danger of
bursting, because one of the thrusts (marked B in Fig. 1 1) has to
be taken entirely by the joint itself. The other three (A, C and D)
are countered by the weight of the pipes or the supports below the
bend. To remedy this, he suggests that when earthenware ppes
are used, one should go to the length of enclosing the whole cdbow
(he calls it a Imee', getiiculus) in a block of red sandstone. This
sounds rather drasdc, but in a *20-digit' pipe (about 4in, 10.16cm
in diameter) with a head of 100ft the thrust at B would be of the
order of 5501b (250kg), and a cemented pipe joint that could
hold together against that would be strong indeed.

This type of closed-pipe system was called a 'stomach' [koilia
in Greek and venter in Latin). In modern textbooks it is usually
called a 'U-bend' or (rather perversely) an ^inverted siphon*.

The other problem — sediment — was faced in various ways,
the most efTecdve being the provision of settling tanks at the
source end. These were long rectangular dstems of stone or con-
crete. The water was fed in at one end, and the rate of traverse
was slow enough to allow most of the sediment to sink to the
bottom by the tune it got to the far end. By taking the oudet from
high up at that end, near the surface, the purest water was drawn
off and fed into the system. Since the tanks had to be drained and
cleaned out at regular intervals they were often built in pairs, and
used alternately. It has already been pointed out that a very

Copyriyhioa inaici lal

46 ENGINEERING IN THE ANCIENT WORLD

serious problem of maintenance would arise if sediment blocked
up the lower part of a U-bend, and a rather cryptic sentence in
Vitruvius (the tesA is almost certainly corrupt) may perhi^ give
a due to the Roman answer* In Chapter 6, he says that coUtviaria
should be bmlt into the U-bend, *to relax the force of the spiritust.
CoUiviaria is a word which is otherwise unknown, and its meaning

Vattey floor

Fig. 11

is obscure. Some editors amend it to colliciaria, which is used else-
where to mean 'gutter-tiles', and interpret the passage as referring
to air-valves, used to release air-locks which might form in the
pipe. A better emendation, however, is colluviaria; colluvium
means sludge, so colluviaria might well have been sludge-cocks (as
they are called nowadays) used to drain the system while repairs
are carried out. Spiritus can mean air, but in a number of other
contexts it is used to mean pcessure (equivalent to the Greek

Copyrighted material

WATER SUPPLIES AND ENGINEERING 47

pneuma)y and this meazung would fit exactly here. Vkruvius says
that when the system is first completed, the water must he fed in
slowly and carefully, or the spifitus will burst the pipes. Unfortu-
nately, he gives no details of the design of the colluviaria, but they
might have been on the lines of the valves used by Hero on his
compressed-air fountain (p. 30).

In view of these problems, it is not surprising that closed-pipe
systems were exceptional in the Greek and Roman world. But
some remains survive of a most impressive one in the Greek city of
Fergamon (now Beigama in Turkey) built during the first half of
the second century B.a, in the reign of King Eumenes XL Accord-
ing to the archaeologists, the source was high up on a hill above
the modem Hagios Geor^^iios. Settling tanks, which probably
stood at the inlet end, were built at a hdght of about 1180ft
(360m) above sea level. From there the pipe ran down to a valley
about 600ft (183 m) below, then rose up about 190ft (58 m) over
a low hill, down again 130ft (40m) into another valley, and then
up again 450 ft (137 m) to the citadel of Pergamon (Fig. 12). If
this was indeed the layout, the pressure at the bottom of the first
bend would have been about 2601b/sq in (18.5kg/cm^).

There is no dear evidence to tell us the material of which the
pipeline was made. If it was earthenware, it seems strange that so
very litde trace survives, as there could be little motive for taking
the pipes away, and they could not have been refused elsewhere.
The surviving stone blocks with circular holes (mentioned above)
would be appropriate for housing the jomts in an earthenware
system. If metal had been used, it might well have been removed
and melted down, and this would account for the lack of remains.
But it is highly doubtful that lead pipes could have withstood the
pressure, and if bronze was used (this hiis been suggested) it must
have been very expensive indeed. The total length was over 3 km,
and though Pergamon at that time had big resources of money,
such a project would have swallowed up a large proportion of the
•gross national product'.

How long this system survived in active use we do not know,
but it clearly did not work satisfactorily in the long term. After the
dty came under Roman government in 133 B.a it was dismantled
and replaced by an open-channel system, carried on arches across
the two valleys and through a tunnd in the hill between them. The
delivery point was quite low down on the slope of the acropolis,

Copyrighted material

1100

iooo

900

SCO

700

€00

500

vertical)

48 ENGINEERING IN THE ANCIENT WORLD

(Sdumatic, horizontal scale much sntatler than

Height
above.

sea level
noo

SettUfifftank

Citadel of
Tergamon

Toint of highest pressure
approK. zeo tb/in^

Fig. 12

and water for the upper town must have been carried up labori-
ously, or perhaps pumped up to the supply points.

Material remains of the pipes and accessories by which water
was distributed from the supply point to consumers are very scanty^
but Vitruvius and Frontinus give us a |ncture of what they must
have been like.

In advanced systems such as the Roman one the supply was
divided into three branches, and three leservoirB were built side
by side. According to Vitruvius^ the central one fed the public

WATER SUPPLIES AND ENGINEERING 49

supply pdnts (of two types — hums, or pools into which one dipped
one's bucket, and saUentes — water-spouts). These were considored
the basic essential, and no other demand was allowed to interfere
with them. The reservoir on one side supplied the public baths,
and that on the other side the private households who had their
own mains supply, and who paid a water-tax {vectigal) the pro-
ceeds of which were used to maintain the public system. Vitruvius
seems to imply that the demand from the central reservoir was
constant, whereas that for the baths, and that for private house-
holds might vary from time to time. If cither of these demands fell
off, the surplus which built up in the corresponding reservoir would
overflow into the public supply reservoir, but the levels were so
arranged that no amount extra demand for baths or pnvatt
customers could interfere with the public supply.

Fnmtinus supplies a lot irf information on the methods by which
supplies were measured and assessed for tax. Here, as in so many
contexts, we meet the characteristic contrast between knowledge
and understanding of the static features, and neglect or ignorance
of the dynamic. No attempt seems to have been made to measure
the speed of flow through a pipe or conduit, and the whole tech-
nique of measurement is based on a calculation of the cross-section
area of a special nozzle which regulated the flow. It is clearly recog-
nized that if the gradient of an aqueduct is steeper, the rate of flow
will be faster, but Frontinus makes no attempt to find out how
much faster. He is concerned merely with a nozde of a specified
size. If the rate of flow is normal, such a nozzle will deliver the
statutory amount of water to the consumer. If the gradient of the
channel is extra steep, or if the supply is increased by heavy rain-
fall in the catchment area, the amount delivered will be above the
legal requirement [exuberare\ but nothing is done about this — it
is simply regarded as a bonus to the consumer. He does speak of
making some adjustment if the rate of flow is slower than normal,
but exactly what he means by 'hghtening' {relevandam) is not
clear.

The nozzle (in Latin, calix) which regulated the supply to a
private consumer was made of bronze, this being a harder metal
than lead, and therefore less easily tampered with. It was about
9 in (22-23 cm) long, its internal diameter carefully measured, and
an dSicial tester's stamp on the outside — at least, it should have
had one. It was set in ^ wall of a reservoir as a rule. It could be

Copyrighted material

50 ENGINEERING IN THE ANCIENT WORLD

let into a pipe or conduit, but this was avoided where possible, as
it was liable to cause leaks. The supply pipe was attached to the
outlet of the nozzle. Frontinus notes that the position of the nozzle
affected the amount drawn off. If set at an angle 'facing the flow*
(Fig. 13) it would obviously collect more, and if slanted the other
way, less. Also, the level of the pipe leading away from the caUx

(a) Plan Aqueduct channel

y y y y yyyy y y j y / j j / j j r r^

Ftow

Ad, cuTsum Square Ad tatus

oppositus (correct conversus

(extra: qmntUy) (U$$ than correct
quantitY) quantity)

(}>) Section

..^^^ Supinus (short)

}AdUbram(comct)
Devexus (extnn)

Fig. 13

affected the rate of flow. If it sloped downwards steeply {devexus)
the flow would be increased. And though Frontinus requires that
the nozzle should be set at right-angles to the flow, and that the
pipe leading from it must be level for some distance, he does not
stipulate a particular depth (e.g. x digits below the surface), which
would have given some kind of consistency to the measurement.
It should be said, however, that the system was fair to the extent
that if the supply was a bit short (owing to droug^) the loss to
each customer would be proportionately the same, or roughly so.

Copyrighted material

WATER SUPPLIES AND ENGINEERING 51

Theoretically there were 24 different sizes of caUx^ but in prac-
tice only 15 were used, and the most important ones are listed in

the Appendix to this chapter. The smallest was called a qftitnana,
'five-to-each-custonier', and had a diameter of about -to in (2.31
cm). This was used as the standard unit of measurement for the
larger nozzle sizes.

For the sake of brevity, let us call this a *no* 5\ There were mx
others between that and the *no. 20', namely nos. 6, 7, 8, 10, 12
and 15. The ^o. 20' was about S^in (8.316cm) in diameter, and
was the standard supply size for a small group of customers. Above
that, there were seven larger sizes in general use, nos. 30, 40, 50,
60, 70, 80 and 90, and the two 'supply main' sizes were the no. 1 00
(centenaria) and no. 120 {centenum vicenum). The sizes of the
smaller and larger nozzles were worked out on two different prin-
ciples. Up to the *no. 20' the number reflected exactly the internal
diameter expressed in quarters of a digit (the digit being -jf o£9l
Roman foot, 0.7275 in or 1.848 cm). Thus the W 5' was i digits
diameter, the *no. 6' | (i.e. If) digits, *no. 8' 2 digits, and so on.
For the *no. 20* and above a difTerent formula applied. The
number denoted the cross-section area of the nozzle expressed in
square digits. Thus the *no. 40' was 40 square digits (1 36.56 cm^)
and the 'no. 100' 100 square digits (341.4cm^). As Frontinus
points out, the two principles of measurement coincide (nearly but
not exactly) at the 'no. 20'. In addition to these complications, the
men in ch^uge of the supply connection and maintenance (Svater-
men', aquarii) had other resources with which to confuse and cheat
their customeis. In some parts of Italy the standard unit was
apparendy the inch {uncia, of a Roman foot, 0.97 in or
2.464cm) instead of the digit. Again, it was not too difficult to
create confusion between a nozzle of I digits diameter with one of
1: square digits area. Even within a highly organized system,
subject to inspection and control, certain dishonest manipulations
were practised. According to Frontinus, in many situations water
was supplied to a reservoir through a 'no. 100' or a 'no. 120', and
the distribution pipes (for each group of — say — a dozen custo-
mers) were 'no. 20'. But the aquarii used a 'no. 100' which was
about 1 3.5 % bigger than it should have been, and a *no. 20' about
20% smaller tl^ the correct theoredcal size. They thus had a
surplus amount in the reservoir, and were able to supply some

Copyrighted material

52 ENGINEERING IN THE ANCIENT WORLD

extra customers illicitly, pocketing the water tax (Tou pays the
money to me, guv'nor, and I squares it up at 'ead office') or at
least a substantial bribe. Since the unofficial customers (again
according to Frontinus) included the proprietors of bawdy-houses,
one can see that the career of an aquarius was beset with manifold
temptations.

There is one simple and obvious question which, surprisingly,
cannot be answered directly from the evidence available. If a
Roman householder had a piped supply of water, did he have a
tap to turn it on and off? The fact that neither Vitruvius nor
Frontinus makes any mention of such a device may merely mean
that they regarded it as too familiar to need description. Hie sug-
gestion mentioned earlier, that the rate of consumption from the
private supply reservoir might vary, cannot prove anything either.
It mi^t shnply mean that more or fewer customers mig^t be
connected up at any one time. If there were no taps, the
water presumably ran fnm a spout into a basin, from which
it flowed away (perhaps being used to flush a lavatory en route)
into one of the drainage channels, and thence into the Tiber.
This was certainly the arrangement at the public water supply
points.

All this gives the impression that the water supply to Rome was
copious, and indeed, by any European standards up to the nine-
teenth century, it was very copious. Exact calculations are im-
possible, because the measurements Frontinus gives are of the
cross-section area of the various aqueduct channels, and even after
Thomas Ashby's meticulous examination of the remains it is
impossible to determine the gradients, and hence make a roug^
guess at the vdodty. However, a conservative estimate comes
out at about 150-200 million gaXkms (680,000-900,000 m"") per
day.

This, however, represents the heyday of water engineering in
the capital city of the Empire, which had financial resources and
expertise available in full measure. In the provinces, small com-
munities had less happy experiences of what was, to them,
advanced technology. A graphic record of one fiasco survives in a
stone inscription* found in tlie Roman town of Saldac (now
Bougie in Algeria, on the coast about 120 miles — 200km — east of

*C(npus Imeri^omm LeHnanm VIII, 2728.

Copyrighted material

WATER SUPPLIES AND ENGINEERING

53

Algiera). It was set up in 152 a j). by Nonius Datus, a retired aimy
surveyor (Ubrator) to celebrate his own achievement in rescuing an
important project from disaster. After telling how he was sum-
moned, at the Emperor's personal request, to take over control of
the work, he says : 'I set out on the journey, and was attacked by
brigands; I and my staff were robbed of everything, and badly
injured. I came to Saldae and met Clemens the Provincial
Governor, who took me to the hill, where they were shaking their
heads and weeping over the tunnel, on the point of giving up the
whole thing, llie tunnellers had covered a distance greater than
that from side to side of the hill !' He explains that the inlet end
of the tunnel, running eastwards, had gone off alignment to the
ri^t (i.e. southwards), and the tunnellers who had started from
the other side had done the same, turning northwards. So, natu-
rally, they had not met in the middle as planned. Nonius Datus
did a fresh survey, much more meticulously this time, and by en-
couraging competition between the two sets of tunnellers, got the
work completed in reasonable time. The inscription ends with a
reference to the great local occasion \vhen the supply was cere-
monially turned on by the Provincial Governor.

Appendix to Chapter 2

THE SIZES OF MEASUREMENT NOZZLES,

AND FRONTINUS' ARITHMETIC

The section of Frontinus' work which deals with nozzle sizes
(1, 23-63) affords us an opportunity to judge the skill and accuracy
in arithmetic shown by an educated but not technically minded
Roman administrator. (On Fix)ntinus' career, see Chapter 9.)

For his calculations he uses a system of fractions. In some places
these are written out, like the whole numbers, in words, while in
the last section (39 onwards) he uses the normal Roman notation
for the whole numbers, and a system of symbob for the fractions.

54 ENGINEERING IN THE ANCIENT WORLD

The system is duodedmal^ and the basic fraction is the unciaj-fxf
indicated by a dash, — .

Thus = means -fi (f),

=5 = — means and so on
S {semis) is used for ^,

so S = — means J + A = f, and so on.
Fractions of a twelfth in general use are the half (24:), written S or
£, the sixth {srxtula, V2) for which there is no symbol, and the
twenty-fourth {tIs) called a scripulus and WTitten 5. (This is the
^scruple' of the old Apothecaries' Weight, one twenty-fourth of
an *oimce', which was one-twelfth of a pound.) For extra
accuracy, Frontinus occasionally subdivides the scripulus into
thirds (1^7). Apart from these basic fractions more or less any
nmnerator or denominator can be used, but they have to be
written out in words — there are no symbols.

In calculating circumferences and areas of circles, Frontinus
takes the value of ir as ^ (3.1428571), which is about 0.04%
higher than the true value. Archimedes had shown three centuries
earlier (in his ticati.sc The Aieasurement of the Circle) that the
true value lies between and ^-f^ , The latter is in fact slightly
closer to the true value (-0.024%), but Archimedes did not esta-
blish that fact, and would obviously be much more awkward to
use as a formula.

Making allowance for this inaccurate value of tr, Frontinus'
arithmetic is quite creditable by any ordinary standards. He says,
for instance (I, 24), that the difference in area between a square
digit and a circle of 1 digit diameter is 1^ of the former and -fr
of the latter. Taking the digit as 1.848cm, 1 square digit is
3.415104cm'. The radius of the circle would be 0.924cm, and
vr* (takuig would be 2.683296. The difference is

3.415104
-2.683296

= 0.731808
3.415104 X A =0.731808

2.683296 X A =0.731808
So the fractions 1^ {tubus quartisdecumis suis) and (tribus
undeciunis suis) though they sound like rough approximations, are
in fact acctirate to at least six places of decimals.

Copyrighted material

WATER SUPPLIES AND ENGINEERING

+ + + + + + + «K +

- I

I

I

o

i

I

i

I

^00 ^

M M

• • • •

o »o

«5 ^
CM O lO <-Ci

o

in

CM CO

•— I r- CT) lOiO lO CO rt<

tO^rfJ^OCNlCM^COc^OCMO)

r

00

+ + + +

CM — I lO
Cs| CM CO CO

t0^0>eo<0c0'*«0r--0 00

co^pio'^epcp^^iooo — cpio
SMMtfiiOdtOOcococbocMcMd)

,+ + + +

H«M tfiCfttn 00 O M C4 <o

«-« ^ f-H l-H I

to

« ca «j

*2 •li^

•S -S I g S I S

(4 (Q C V V C d

i

< « < «

< CQ < ffl

IOOO(NCMOOQQQOQOO

m ~^ S ^ ta

Si

•M

1^

56 ENGINEERING IN THE ANCIENT WORLD

The most common sizes cxf calix, as calculated by Frontinus in
I 39-83, are listed in Table 2. For the sizes up to no. 20 he starts
from the diameter in quarter-digits, as explained on p. 51. From
this he works out the circumference (col. 3) and the capacity in
quinariae (col. 7), this being the ratio of the cross-section area to
that of a quinaria nozzle, IJ digits diameter (4.191 cm^). A spot
check on a few of his figures, chosen at random, gives an indica-
tion of his accuracy.

Take his figures for the 'no. 12A' nozzle. The diameter is 3
digits (5.544cm). He gives the circumference as 9+ Ig-I digits
(9.4270833) and the true value is 9.4285714. The difference is
0.001 488 1 (or, in his terms, less than one-third of a scripulus) and
the error is approximately —.015%. The capacity in qidnariae
he gives as 5 + fif (5.760416 6), while the correct figure
is 5.7599995. The error is 0.0004171 of a quinaria, or
+0.007 2%.

For the sizes above no. 20 he starts from the figure in col. 5 and
works back to col. 1, which involves extracting a square root, e.g.
for the 'no. 80', d = 2 V^*^. His answer, 10-h (10.090 277) re-
quires something better than 1 figure tables to correct it. Checking
back from that figure to column 5 (still taking tt = ^) gives a
figure of 79.996 468 square digits, an error of —0.004 4%, which
has been compounded by squaring. From the figure in col. 1 the
circumference works out at 31.712 298; Frontinus gpives 314-ii
(31.708 333), an error of -0.012 5%. The figure for quinariae in
col. 7 is 65+^ (65.1666), and error of 0.003 oi a quinaria^ or
-0.0046%.

The exceptions to this generally good standard of accuracy are
the figures given for azes *as used hy the watermen' — 12B, 20B,

lOOB and 120B. It may be significant that no figures are given for
the circumferences (this being the crucial measurement in the
manufacture of the calix), and the other figures, with two excep-
tions, are all in round numbers (4^, 12 and 16 digits, 6, 13 and 92
quinariae). The remaining two figures can be reasonably explained.
The 'watermen's no. 1 2' was supposed to deliver 6 quinariae, as
against the 5+ fit of Frontinus' calculation (no. 12 A), and ac-
cordingly, the diameter is corrected by which is within a very
narrow margin of the correct amount. The quinariae for the
Vatermens' no. 120' present a more difficult problem. It looks as
though Frontinus has worked it out properly, instead of accepting

Copyriyhioa inaici lal

WATER SUPPLIES AND ENGINEERING 57

a round figure (the watermen probably reckoned it as I64c\ but
he goes on to say that it is twice the capacity of a no. 100 — that is,
his own theoretical lOOA. In this he is a lA^iole quinaria out, and
unless one assumes some alteration in the text, it must be admitted

that here, for once, he has been a little careless.

Copyrighted material

3

Water pumps

In the ancient world, pumps and similar devices were used for
various purposes. Much the most important was the irrigation of
crops, especially in market-gardens, but they were also used for
pumping flood water out of mines, bilge>watcr out of ships (the
larger and better equipped merchantmen) and, according to Hero
of Alexandria, in fire-fighting equipment

The simplest and crudest dates from the very remote past, and
was in use all over the Middle East long before ibt dasstcal period,
and illustrations of it are found on Greek vases of the nxth century
B.c."^ from the time when work scenes first begin to appear as
subjects in vase-paintings. Nowadays il is usually called a swing-
beam or swipe, or by its Arabic name shadouf. The Greek name for
it was keldneion, and the Latin names ciconia ('stork', from its
resemblance to that bird) and tolleno ('lifter'). It consists of a
support, usually a tree-branch with a fork at the top, driven into
the ground about 8-lOft (3 m) away from the well-head or other
source. This acts as a pivot for a beam, one end of which is direcdy
above the well-head, and has a bucket suspended from it on a
rope. Towards the odier end is a counterweight, commonly a large
stone, which is shifted along the beam until it just out-weighs the
bucket full of water. The rope is pulled down until the bucket
dips into the water and fills. It is dien released, and the counter^
weight lifts the bucket a few feet above the ground, where it is
emptied by hand into another receptacle, or into a conduit <rf
some sort.

Using one of these machines is not much less laborious than
using a bucket on a rope — it is a little easier to reach up and pull
downwards than to reach down and pull upwards. In an amusing
passage in Menander's comedy, The Misanthrope (fiyskolos) a
wealthy young man who, for romandc reasons, has been uang a

*e.g. Atdc Mack-figure vase in Berlin, rqaoduced in £. Pftihl, Makni
midZ«ehmmg der Cntehen (Munich, 1923), vol. Ill, pi. 72, no. 276.

Copyrighted material

WATER PUMPS

59

mattock^ describes how his back, shoulders and hips all locked
solid from the unaccustomed hard work, untO he was 'swinging up
and down in one piece like a shadouf . It was clearly a familiar

sight to the Greeks and Romans, but was so simple and obvious
that the writers on machinery did not consider it worth mentioning.

In many situations, water for irrigation had to be raised from a
stream or canal, over a low bank and onto an adjoining field. The
quantity required was large, but the head lift only a few feet. For
this purpose two types of pump were used, the screw and the drum.
The Greek word mechane ('machine') is frequently used in con-
texts connected with irrigation, and might refer to cither type, and
it is worth noting that the need for such a device was so widespread
in Egypt that in Greek documents from Qxyrhynchus a piece of
cultivated land comes to be called *a m&kane\

Fig. 14. Archimedean screw pump.

Where a screw pump is specified, it is called cochUas in Greek

and cochlea in Latin, from its resemblance to a spiral seansbell*.

Tradition ascribed its invention to Archimedes in the third century

B.G., and though this may be true, it cannot be relied upon

absolutely — the pump might have been in use earlier. Vitnivius

gives a detailed description of how it was made (X, 6). The

rotor is made from a round wooden pole, its diameter -iVth of its

length (Fig. 14). The circumference is divided into eight equal

segments, maiked by parallel lines along its length. Then the

length is divided into sections, each equal to one-eighth of the

*The modem Egyptian version in the Science Museum in London is
constructed quite diiferendy.

Copyrighted material

60 ENGINEERING IN THE ANCIENT WORLD

circumference, and marked by rings drawn around the circum-
ference. The dff ect of this is to produce a pattern of small squares
all over the surface, and since the blades are placed diagonally
across these squares, the pitch of the spiral is 45^.

The blades are constructed by taldng a flat strip of willow or
osier, coating it with pitch and placing its end on one of the longi-
tudinal lines at one end of the shaft. Then it is drawn obliquely
around the rotor shaft and nailed onto it at the intersection of the
two lines, ^th of the circumference along and the same distance
around the shaft. It is then taken along the same diagonal course
and fixed over the series of intersections, returning at the eighth
point to the longitudinal line from which it started, having made
one complete circuit of the shaft. Then it continues along the
same padi to the far end of the shaft, making just over five circuits
altogether. This f mns the base of one blade, and more strips are
ptched and nailed on top of it until the diameter of the rotor is
built up to twice the original diameter of the shaft (i.e. ^di of
its length). Vitruvius' account suggests that seven more blades

were then positioned by the same method, one starting from each
of the longitudinal lines. TTiis is by no means impossible, but it
raises some problems, which will be dealt with later.

A wooden cylindrical case is then made to fit around the blades
(however many), constructed like a barrel, the planks painted v^th
pitch and bound with iron hoops. There is virtually no doubt that
in Vitru\ ius' design the case was fixed to the rotor blades. There
is some obscurity in his description of the way in which the rotor
was mounted, but it is dear that it had an iron cap <»i each end
and (probably) an iron spigot which turned in an iron socket on
each of the supporting posts. The rotor with its case was turned
'by men treading' {hominibtu cakaniibus) but exactly how this
was done is not clear. Vitruvius recommends that the rotor should
slope upwards at an angle of about 37°. Though this is arrived at
from the well-known construction of a right-angled triangle with
its sides in the ratio 5 :4:3, it probably represents an approximation
to an optimum angle which has been arrived at empirically. (*Us
put 'un up like this-yur, an' 'ee' wurked allright ; us put 'un higher
an' 'ee didden wurk so gude, so us put 'un back where 'ee be, an'
let 'un bide'.)

How, then, was it turned ? Vitruvius ja^ovided a diagram in his
original wotk, but this has been lost, and the drawing? which

WATER PUMPS

61

appear in our earliest manuscripts of the De Architectura are the
work of medieval scribes and commentators. Most of them show a
treadmill mounted outside the case (clearly, they assumed that it
was fixed to the rotor) rather improbably placed halfway down and
tilted at the same angle of 37°. Even if it were in a more con-
venient position, it would still be extremely difficult to tread a
wheel tilted at such an angle and, needless to say, very inefficient.
A terracotta in the British Museum* shows a workman steadying
himself on a crossbar at about waist height, and turning a cylin-
drical object with his feet by treading a row of cleats around its
middle. The fact that the cylinder tilts upwards towards the right
suggests that it might be a screw pump, but the dlt is very much
less than the angle Vitruvius suggests, and the modelling is crude
and perhaps inaccurate. Another ancient illustration is to be found
in a wall-painting at Pompciif. This shows a slave standing in the
shade of a small structure like a lych-gate, turning a compact-
looking cylinder with his feet. It is generally taken to be a screw
pump, but this may not be correct. The output is pouring into
what looks like a storage jar buried in the ground, so it might
possibly be a mill or crusher of some kind. The angle of tilt is, if
anything, even less than in the terracotta, and the lift appears to
be no more than a few inches.

Some remains of Roman screw pumps and their mountings
have been found in Spanish mines, and they appear to have been
tilted at an angle of about 15^. Thus there is evidence for a variety
of different indinadons, and three comments should be made.
Firstly, the best angle for efiGidency is dearly related to the pitch
of the spiral, and this was no doubt taken into account when
designing the screw. Secondly, for a given pitch of screw, the
amount of water lifted by each blade is reduced if the pump is
tilted to a higher angle, and eventually drops to zero when the tilt
becomes equal to the pitch. Vitruvius' pump would cease to work
altogether above 45°, and the angle he suggests (37°) was appar-
ently chosen to give the maximum possible lift at the cost of
reduced output. A different compromise would be reached if

♦3rd-2nd Cent. B.C. Illustrated in Henry Hodges, Technology in ^
Ancient World (Pelican, 1971 fig. 211, p. 184).

fin the Gasa dell' Eiebo. Illustrated in Rostovzeff, Social & Economic
History of the Roman Empire (second ed. O.U.P. 1957) pi. LIII, 5 and
R. J. Forbes, StiuSes in AnaaA Ttchnobgy (Leiden, 1963) vol VII, p. 213.

62 £NGIN££RING IN THE ANGI£NT WORLD

output were more important, and lift mattered less. Incidentally,
the device was, very roughly speaking, self-compensating. As in-
creased tih increased the head but decreased the output, the work
done at any angle, and hence the power required, would remain
approximately constant for a given speed of rotation. Thirdly, the
number of blades that could be mounted on the rotor might be
limited by the angle of tilt. If the head is low, the quantity of water
lifted by each blade may be curtailed if the next blade above it is
too close. Here again, Vitruvius' design is a reasonable compro-
mise. If, for instance, the inside diameter of the pump case was
about 1ft (30cm), the water above each blade would extend about
7.4 cm up the inside of the casing. With all eight blades mounted
on the shaft (as Vitruvius seems to suggest) the spacing between
them would be about 5.9cm, and only a minimal reduction in the
amount of water lifted by each blade would result.

Surviving remains of rotors from screw-pumps do not appar-
ently have as many as eight blades. One of the most interesting, in
the Liverpool Institute of iVichaeology,* has tw^o, starting from

diametrically opposite points, and the pitch is about 15°. One
explanation might be that it was intended to work at a lower
angle of tilt. The addition of extra blades would, of course, increase
the output per revolution, but it would add appreciably to the
weight and the cost of the pump, and would be pointless if the
available driving power could be fully taken up with fewer blades.
Finally, one very significant advance in design is evidenced by a
surviving screw found in the Centenillo mines near Linares. Its
blades were made from sheet copper, ^in (3.2 mm) in thickness,
and were fixed to the rotor and to the case by small metal brackets.

To return to the question of the angle of tilt, there is one more
observation which may be relevant. Vitruvius describes, as though
it were something not particularly remarkable, a right-angle drive
in the water-mill. It is quite clear from this that the engineers of
his day were capable of devising gears w'hich could have coupled
a horizontal treadmill shaft to a tilted pump shaft, but there is no
evidence to show that this was actually done in antiquity. What-
ever the arrangement may have been, the man or men turning
the pump used a treading action, and it is an interesting commen-
tary on human progress that the photographs of screw ptmips in

♦Illustrated in T. A. Rickard, The Mining of the Romans in iS^pum, J.R.S.
VIII (1928) 129-143, and Plate XII/1.

Copyrighted material

WATER PUMPS

63

use in Egypt today almost all show two men seated on the ground,
one holding the mounting steady and the other turning the rotor
with a handle. As anyone knows who has cranked a reluctant car
engine on a winter morning this is a much less efficient and much

more back-breaking way of harnessing human power.

It would be most interesting to reconstruct a pump of this kind
to Vitruvius' specification, test it at various angles of tilt, and make
an assessment of its efficiency. But the best that can be offered here
is a very approximate guess at its performance, based on a number
of assumptions which are arbitrary, and may prove to have been
incorrect. Suppose the rotor to be nearly 8ft (2.4m) in length, and
nearly 1ft (0.3 m) in diameter. If mounted as Vitruvius suggests
this would raise the water to a head of nearly 4ft (1.16m).
Given an available power, from one 'man treading', of 0.1 h.p.,
the amounts pump^ at various rates of efficiency would be as
follows :

at 60% effidency just over 50 gall (about 235/) per minute
at 50% „ just over 40 gall (nearly 200^ „
at 40% „ nearly 35 gall (nearly 1601) „

There would be appreciable loss of energy through friction in the
rotor shaft bearings, and also through spillage of water over the
rotor shaft due to uneven motion of the rotor, but it would be
reasonable to guess that the efficiency might be somewhere in the
region of 40-50%. Even so, the ouq>ut is by no means contemp-
tible. 2,000 gallons of water in an hour could make a vast differ-
ence to a parched vegetable garden.

The other type of low-lift, high output pump described by
\^truvius (X, 4) was called tympanon in Greek (spelt tympanum
in Latin), the name meaning 'drum', and is simpler to construct
than the cochlea. It has a horizontal axle, turning in bearings

supported on posts or beams at each end. X'^itruviiis recommends

that both the axle and the bearings should be 'plated \vith iron',
but this may not have been the usual practice. A few Greek
papyrus documents from Egypt refer to the supply and collection
of replacement wooden axles (and, incidentally, the *handing-in'
of the old ones), from which it is dear that wear on the axle was
a major problon. On this axle turned the drum itself, made from
planks in the form of two parallel discs, joined together by
dght partitions which divided the drum radially into ei^^t

Copyrighted material

64 ENGINEERING IN THE ANCIENT WORLD

compartments (Fig. 15). Hence the pump is sometimes referred to
as a 'compartment wheel'. The sides and partitions were all sealed
with pitch 'like a ship's huir, and the outside rim was also boarded

in, except for a slot about Gin (15cm) wide, opening into each
compartment at the end which was to enter and leave the water
first as the drum revolved. A circle of eight holes was cut in one
side of the drum near the axle, one opening into each compart-
ment, to act as outlets for the pump, and a wooden trough (known

Fig. 15

as a 'launder') was placed so as to catch most of the water as it
flowed out, and channel it into a conduit. The drum, like the screw
pump, was turned by 'men treading'. If the drum itself was made
to act as a treadmill by fixing cleats around its rim, care would
have to be taken to avoid excessive resistance as the cleats passed
under water. There may have been a separate treadmill mounted
on the same axle, though Vitruvius does not specifically say so. In
any case, the axle was horizontal, and the problem of tilting did
not arise.

The design of this pump is very ample and reliable, and the
only loss of energy (though it might be serious) occurs in the asde
bearings and in the 'paddle effect* of the cleats passing through
the water. But it has a number of limitations. Firstly, the absolute

WATER PUMPS

65

maximum height to which it can raise water is a little less than
half its own diameter, and in order to raise a significant quantity,
the head must be reduced to something like two-thirds of that
height. This means, for instance, that a drum nearly 10ft (3 m) in

diameter would not have a very effective output at a head of more
than 39 in (1 m). If a drum of this diameter had a thickness, mea-
sured internally, of nearly 8 in (20cm) each compartment would
scoop up about 11 gall (50/) of water, giving an output of about
88 gall (400 Q per revolution. A very ^proximate idea of the
pump in action can be obtained by supposing that the drum
revolved at 2 rpm, which would produce an output of about
1 76 gall (800Q per minute. To turn it at this speed would require
at least two men working quite hard (0.2 h.p., widi 20% Mction
loss) and it might not be posdble for them to keep it up over a
long period. To make the picture more precise, let us imagine the
drum itself forming the treadmill, with cleats around its circum-
ference at intervals of just under 10 in (25 cm). There would
be 38 treads altogether, and to turn the wheel at 2 rpm the men
would have to tread 76 times per minute. It would feel like climb-
ing a steep flight of stairs at a brisk walking pace.

In addition to the limitation on the head lift, the drum has two
others which the screw pimip does not have to the same extent.
One is that the output has to flow through the holes near the axle.
No dimensions for these are given by Vitruvius, but his term for
them, columbaria ('dovecote') suggests that they were not large,
and Uieir width is in any case limited by the tapering of the com-
partments towards the centre. As a result, there is a limitation on
the output of the pump which is reached when the speed of revo-
lution leaves insufficient time for the compartments to empty as
they pass over the axle. Beyond that point increasing the applied
power, and the speed of rotation, will not bring about a propor-
tionate increa-se in output. But with the screw pump the speed and

output can be increased indefinitely — until, that is, the casing
bursts or the lower support disappears into the mud. It is unlikely,
however, that there would ever be sufficient power available to
turn the drum fast enough to encounter this limitation, except in
unusual emergencies, such as sudden flooding of a mine-shaft, or
desperate renewal of supplies for irrigation after a long drought.

The third limitalion must have been encountered quite often
when raising water from a muddy ditch or canal. There is a

uopyiiyhiea maiciial

66 ENGINEERING IN THE ANCIENT WORLD

tendency for mud and debris to pack around the outlet holes and
reduce the flow through them. Because the rotor of the screw
pump presents, m effect, a smooth channel without sudden bends»
it is capable of pumping slurry widi quite a considerable solid
content — though there comes a time, of course, when it eventu-
ally clogs up. This is why the screw pump is still in use nowadays
for such purposes as pumping untreated sewage, or clearing
animal refuse from farm buildings. It is usually called an 'auger',
and is made with a fixed case and a steel rotor.

In conclusion, a word should be said about the comparative
efficiency of the screw pump and the drum. In the figures for per-
formance given above — which, it cannot be too often stressed,
are mere guesswork — it has been assumed that the drum was
rather more efficient than the screw. As against this, ancient
writers in more than one context comment on the remarkable
efficiency of the screw. Two examples wiU suffice. Diodora
Siculus (V, 37, 3-4) remarks that a surprising quantity of flood
water was pumped out of Spanish mines by this means, and in a

description of a merchant ship built in the third century B.C.
(Athenaeus V, 208 F) which was enormous by the standards of
the time, it is explicitly stated that one man was able to do all the
bilge-pumping required by means of a screw pump. It must be
said, however, that this claim should be regarded with suspicion.
From a rough estimate of the overall size (based on the cargo
capacity) we may infer that the ship's draught was in the region
of 15ft (4.6m), and a single screw-pump would hardly have given
the necessary lift.

Where it was necessary to raise water to a greater height, a
number of other devices were used. All of them sufifered from ^e
drawback of a limited output, and where both a high lift and a
large output were required, the only course open was to construct
a series of pumps feeding each other. This must have been very
costly to build and operate, and only in a situation such as silver
mining, where increased output of ore might offset the costs,
would it have been worth while. In agriculture this would hardly
ever be so.

The simplest of the higher-head devices was the bucket-wheel,
a modification (or possibly a precursor) of the drum, described
very sketchily by Vitruvius later in the same chapter (X^ 4). The
Greek name for it was polykadia (^ulti-bucket*), but \^tnivius

Copyriyhioa inaici lal

WATER PUMPS

67

does not give a Latin name. It conasts of a wooden wheel with
buckets fixed around its rim, and its axle mounted at such a hei^^t
as to allow the buckets to dip just below the surface of the water

to be lifted as the wheel is turned. Each bucket scoops up a quan-
tity of water, and as it passes over the top of its orbit its contents
are tipped into a wooden trough. The shape and positioning of
the buckets is clearly very important. Vitruvius calls them modioli
quadrati (rectangular grain-measures), which perhaps indicates
that they were broad at the base and tapered to a narrow opening
at the top since this was the usual shape for such a measure. He
also says, 'when they have reached the high pomt and are return^
ing downwards, they pour their contents into a reservoir . .
which suggests that they were tilted at such an angle that thdr
contents did not b^in to pour out until they had passed *top dead
centre'. If this was in fact achieved, it represents a big improve-
ment on the round earthenware pots still to be seen on wheels of
this type in some Mediterranean countries.

The design w^as forgotten, and re-discovered in the Middle
Ages, since when it has been known as a 'Persian wheel'. Nowa-
days these wheels are usually turned by animals. Though Vitruvius
makes no mention of this, there are dear indications in papyrus
documents that oxen or donkeys were used in Egypt by the Greek-
speaking communides in the second century a j>., if not earlier. In
one document, the writer excuses himself for not fulfilling an irri-
gation contract 'because there was not enough fodder for the
animals'. In another, arrangements are made for supplying water
to *thc men who drive the animals round*, and another deals with
the supply and purchase of 'connecting straps' {zeukteriae) for
harnessing the animals. In all these documents the pump is called
simply mechaniy which could mean any type but most probably
means either a *druin' or a bucket-w^heel. Since each of these
requires a horizontal drive, and the animals roiild only have turned
a capstan with vertical axle, some form of gearing must have been
fitted, either a bevel gear or — more probably — a primitive crown
wheel and pinion. (The only possible alternative might be a tread-
mill big enough for a donkey to walk around inside. Such an
apparatus was used for hauling up water from a deep well at Greys
Ciourt, near Henley, until the First World War, and can still be
seen there.) In another ancient document there is an account of
a fracas in which some parts of a pump were set on fire and partiy

Copyrighted material

68 ENGINEERING IN THE ANCIENT WORLD

destroyed. The word used — er gated — might mean the bars of the
capstan, or (posfflbly) the crown wheel and pinion.

The obvious advantage of this type of pump over the drum is
that a wheel of the same diameter can raise water twice as high —
perhaps even more, but it has one disadvantage which the drum
does not share. As the wheel revolves, the buckets are tilting all
the time, and there is a danger tliat some of their contents will be

Fig. 16

spilt, either before they reach the launder, or after they have
passed it, or both. Vitruvius may have provided one solution by
choice of shape for the buckets. Another ingenious one can be
seen from the remains of a Roman pump of this type found
(during the early 1920's) in the Northern workings of the Rio
Tinto mine in Spain. The wheel itself^ 14ft lOin (just over 4.5m)
in diameter, was made entirely of wood — the hubs and bearing
of oak, and the rest of jnne — and wooden doweb were used
throughout the construction, presumably because iron nails would
have rusted too rapidly in the wet conditions. The axle was made
of bronze, square in section with its ends rounded where they
fitted into the bearings. There were 24 spokes running alternately
from each hub to the inside of the rim, which was made hollow
and rectangular in cross-section (as shown in Fig. 16), the internal
measurements being about 7 X 5in (18 X 13 cm}. The whole
densadon — the olive oil would make these more easily visible. If

Copyrighted material

WATER PUMPS

69

The openings through vMch the water entered each compart-
ment haai the sump and left it at the top of the circuit were on
the side of the wheel, and roughly quadrant-shaped, their strai^t
edges being on the inner rim and the leading end of the compart-
ment (see Fig. 16). This arrangement ensures minimum spillage
before, and complete emptying of the compartments after, 'top
dead centre'. The water was collected from the outlets in a launder,
which was fixed as close as possible beside the wheel without rub-
bing on it, and as high as could be managed without serious
spillage. Palmer's estimate of 25 % loss seems rather pessimistic,
but may not be far out. Combined with power losses from friction
in the axle bearing9 and the paddle effect of the rim passing
through the water in the sump, this brought tiie efficiency of his
reconstructed model down to just over 60% . Its output was nearly
19 gall (86.4/) about 19 gall per minute, assuming a theoretical lift
of 12 ft (3.65 m). The diameter of the wheel was more than this,
but when allowance is made for the width of the rim dipping
under water, and for the launder being low enough to catch
most of the outflow, the actual lift was probably about equal to
this theoretical figure. The power required to work this wheel —
just over 0.1 h.p. — could have been provided by one man over
an 8 hour shift

One of the most ambitious Roman mine-draining systems, the
remains of which were found in another part of the Rio Tinto
mines, had a succesaon of eight pairs of these wheels, and was
designed to raise the water altogether about 97 ft (29.6 m). The
arrai^ement is quite ingenious. There is a natural tendency for
the wheel to propel the water along the launder, and the flow
is in any case pulsating rather than steady. If two wheels were
discharging into parallel launders, there would be a tendency for
turbulence to be set up in the region where the launders joined,
which could only be corrected by increasing the downward slope
of the conduit leading to the sump for the next higher pair of
wheels, and this in turn would involve losing a few inches of lift.
So the two wheels (side by side) are made to revolve in opposite
directions, and a single launder runs around both (Fig. 17), so
that each wheel propels die water along it in the same direction,
and only a very slight fall is necessary. This system would have an
impressive output — about 2,400 gallons per hour at a head of
nearly 100ft is no mean achievement, but it should be remembered

70 ENGINEERING IN THE ANCIENT WORLD

that the total cost must have been considerable, and that it needed
the full-time labour of 16 able-bodied men* The ore deposits must
have been very rich to justify such an investment.*

The second high-lift device is a ample and logical extension of
the bucket-wheel, and has the great advantage that the height to
which it can raise water it not limited by the diameter of the vfhed,
but only by the available power, in relation to the quantity re-

♦A reconstructed part of one of these wheels can be seen in the 'Roman
Life' room at the British Museum. A remarkable feature is that they were
apparently made up from 'construction kits', the wood for some of the
parts not being native to that part of Spain, but imported from elsewhere.
For ease of construction (in the place where they were to be used) some
of the side pieces for the rim-buckets were numbored in Roman numerals.
They have been wrongly rec<mstructed in the British Museum. It is quite
dear that the pieces would be consistently numbered either on the outside
or on the inside, and as a result of ignoring this, the buckets have been
reconstructed with outlet holes on both sides of the wheel. The rate of
outflow through one side would have been ample, and a second outlet
would have required a second launder, and would have doubled the
wastage by spilling. The method of constructing the hub and of fixing the
spokes in it, can be seen quite clearly. The same error is repeated in the
drawing of the Dolau Gothi wheel — see G. C. Boon and C. WilUams in

Fig. 17

JRS LVI (1966), 122-7.

Copyrighted material

WATER PUMPS

71

quired. Vitruvius gives a very brief sketch of it at the end of the;
same chapter (X, 4) without givmg it a name, but from the
passing reference in Hero (p. 207) wie gather that its Greek
name was halym ('chain'). It had a treadmill on a horizontal axle,

and two parallel endless chains were suspended from the axle,
with buckets fixed to them at intervals. According to Vitruvius,

Fig. 18

the standard size of bucket was one congius (about 5^ pints or
3.3/). The chains were long enough to hang down to the water
level below, so that the buckets dipped in as they reached their
lowest point. They were then held upright all the way up to the
axle, tipping over automatically as they reached it, and emptying
their contents into a conduit (Fig. 18). This arrangement can be
made to work more efficiendy than on the bucket- wheel, since the
buckets do not begin to tip until they actually reach the axle, and
then they turn over rapidly and decisively. Waste by sfnllage can
be reduced to a negligible minimum.

There is, however, one serious mechanical problem which must
have arisen on this madiine. Unless the weight of water being

Copyrighted material

72 ENGINEERING IN THE ANCIENT WORLD

lifted at any one time (Le. the contents of half the buckets) was
small in relation to the total weight of the chains and buckets,
which in turn means that the output was small, there must have
been a tendency for the chains to slip around the axle. At first

sight it might appear that Vitru\aus was unaware of this problem,
and has neglected to describe some sort of sprocket arrangement
on the axle, to engage with alternate links of the chains. This
need not have been anything more sophisticated than a set of
headless nails, driven into the axle at the appropriate spacing.
There are, however, two possible ways in which the problem may
have been soKed. Spealdng elsewhere of a rope pasmng over a
pulley, Vitruvius uses the word traicere. Here, however, he says
that the chains are Svrapped around' the axle (the word he uses
is involvere, an exact Latin equivalent for the Greek peri-eileo,
which Hero uses of a cord wound several times around a spindle).
This presumably means that the chains passed at least 1^ times
around the axle, which might give sufficient grip to prevent them
from slipping much, especially if the axle was left rough, or fluted
at about the right spacing for the chain links. With this arrange-
ment in operation, as each bucket came round for the second
time it would push apart the rising chains as diey met the axle,
hold them apart for the next half-turn, and then allow them to
return to their normal spacing as the next bucket approached —
not, of course, without some friction and loss of energy. This
arrangement would also limit the number of buckets, since the
minimum spacing between them would have to be about seven
times the diameter of the shaft, to make certain that each one
cleared the shaft and allowed time for the chains to re-adjust
themselves before the next one arrived. If anyone should ask why
the (to us) obvious alternative — some sort of qirocket arrange-
ment — is not mentioned, the answer could lie in the quality of
iron available to the chain-maker. Unless made with very heavy
links (which would be expensive) the chains would inevitably
stretch in use, and after a short time the sprockets would cease to
engage properly. In the modern world, this disaster comes to those
who ride ancient bicycles with chains that have suffered too much
for too long.

But there are a number of objections to this hypothesis. How-
ever Utde the chauis may have slipped, it is most unlikely that each

WATER PUMPS

73

would slip by the same amount, so, after a time, the buckets would
hang askew, spiU their contents, and jam on the shaft. But Fhilo^s
account of the repeater catapult offers a poflsible alternative. The
shaft may have been faceted — in the form of a pentagon or hexa-
gon, and the chain made with straps (not links) to fit the facets.
The buckets could have been suspended from horizontal rods
fixed between corresponding straps (see Fig. 19). This would
eliminate slipping altogether, and ensure that the buckets hung
level. It would also eliminate the problem of power-drive on the
lower shaft.

Fig. 19

The bucket-chain must almost certainly have been more expen-
sive to construct than the bucket-wheel (which could be made
without ironwork), and only in situations where the wheel could
not be used would the chain be preferred. Two such situations
would be (a) when the total lift had to be more than about 20ft
(6m) — it would be difficult to make a wheel bigger than that in
diameter — or (b) where the source of water was a well or other

Cc, , y od material

74 ENGINEERING IN THE ANCIENT WORLD

place of restricted access, where it would be impoflsibk to dig a pit
big enough for a wheeL

Let us try to imagine a bucket-chain in actkm — once again, by
conjecture based on a number of arbitrary assumptions. Suppose
the chains to be just over 1 10ft (33.6m) long, Svrapped around'
a shaft 15|in (40cm) in diameter. The maximum possible lift,
with the buckets dipping just below the water, would be about
52 ft (slightly less than 16m). On chains of this length, twelve
buckets could be attached at intervals of just over 9ft (2.8m); if
the buckets were of the size recommended by Vitruvius (one
congius), one complete revolution of the chains would raise just
over 8f gall (39.6 Q of water. To accomplish this, the axle would
have to be rotated about 27 times which, unless the treadmill was
quite small, might well take as long as three minutes. At that speed
the output of just under 3 gall (13.2i) per minute would represent
one man^s work at an efficiency of just under 50% . The impUcap
tions of these dramadc, but by no means fanc^ figures, are
clear. It would be impossible to irrigate more than a tiny area of
cultivated land by this means, and it is most unlikely that the
produce of such an area would balance, or come anywhere near
to balancing, the cost of installing and running the macliinery.
On the other hand, it might supply the domestic needs of a
modest villa — drinking, cooking, washing and sanitation — even
if these had to be met from a well rather deeper than usual. If
worked for two hours a day it would provide 350 gall (nearly
1600/).

There is a means of escape from these severe limitations where
the source of water is a river or fast-flowing stream, which can
provide an alternative to man-power. At the start of the next

chapter (X, 5) Vitruvius describes how a water-wheel was used
to work a bucket-chain. It was apparently a paddle-wheel of the
under-shot type, which, if efficiently constructed might well
deliver more power than one man. A few extra problems might
arise; the power was applied at the bottom of the chain circuit
instead of the top, and whereas the weight of the chains and
buckets provides extra grip on the top axle, it would not help
here. In fact, it is almost certain that the faceted shaft and
strap-chain were used. But the great advantage of this system
would be tiiat it could be kept running 24 hours a day, without
employing any able-bodied men. An old or infirm skve could

Copyrighted material

WATER PUMPS

75

keep an eye on it, and raise the alann if anything went wrong.
Aasuming that the water-wheel developed three man-power'
(0.3 h.p.) and ran condnuously, it could perhaps deliver 12,600

gall (54,000/) per day.

In what situations would it be desirable to raise a comparatively
small amount of water from a running stream to a high head?
There are two which spring to mind. One is a villa built on rising
ground near a river, with domestic supplies pumped up by one of
these machines, and perhaps reservoirs (stagna) to be kept fiDed.
The other is where a river fkiws across a plain, and it is necessary
to provide irrigation some distance away from the nearest point
on the river. Using the machine described above, water could be
raised to the top of a tower on the river bank and taken thence by
a sloping conduit (or rather, a miniature aqueduct). With a
gradient of 1 in 200 (not uncommon in aqueducts) the water
could be conveyed over a distance of nearly 2 miles (about 3.2km)
to the land requiring irrigation.

Finally, there was a fifth type of pump for which we have good
literary evidence and some archaelogical remains. This Is the
force-pump, wdth pistons, cylinders and valves, called in antiquity
a *Ctesibian machine', after its inventor, Ctesibius (third century
B.C.). A description of this type of pump is given by Vitruvius
(X, 7) and by Hero of Alexandria, whose fire-engine made use of
it {Pneumatic a I, 28). Both these writers also give descriptions of
an organ, in which a jnston-and-cylindcr pump was used to force
air into the reservoir {PneunuUica I. 42, \^truviu8 X, 8) Hie
pneumatikon arganon referred to by Pliny {Nat. Hist. 19, 20), for
watering vegetable gardens from a well, was almost certainly a
pump of this type.

The design was as follows. There were two vertical cylinders,
their pistons worked reciprocally l)y a rocker-arm (Fig. 20). If
this was accessible at ground level (as in the fire engine), a handle
was fitted to an extension of the rocker-arm, at one end or both
ends. If the pump was submerged, or down a well, a wooden push-
rod must have been attached to one end of the rocker-arm. The
connecting rods pivoted on the rocker-arm at their 'big ends', and
it is interesting to note that Hero says {d propos the OTgan pump,
at the very end of Chapter 42) that it is better to have a pivot at
the piston end of the rod, to avoid lateral motion of the piston.

Copyrighted material

76 ENGINEERING IN THE ANCIENT WORLD

This pivot corraponds exactly to the little end and gudgeon
pin in a modem petrol engine, and would help greatly to reduce
wear on the piston and cylinder, eq>eciaUy when the connecting
rods were short In the absence oi any flexible tube or connection,
the cylinders must have been rigidly mounted.

Fig. 20. Ctesibius'i water pump.

Both Hero and Vitruvius specify that the pistons and cylinders
should be made of bronze, this being probably the strongest metal
which could be worked to the necessary degree of predaon with
the took then available. It would be very nice to know more details
of the manufacturing process. Hero mordy says that the cylinders
liave thdr inner surfaces worked to (fit) the pistons, so that there
is no by-pass*, and that this was done on a lathe. Vitnivius uses
the phrase oleo subacH, which is difficult to interpret Subactus is
used elsewhere to mean precision-turned on a lathe. If that is the
meaning here, then oleo must mean that olive oil was used to
lubricate the cutting tool. But the more natural meaning of
subactus would be 'forced', which might mean that the final
trimming was done by ramming the piston into the cylinder, using
olive oil as a lubricant so that each was 'trued up' by the other.
The technique is known as 'lapping'. All this, however, is very
speculative, and it should be added that olive oil is a poor lubri-
cant for metal surfaces and would do little to ease the movement

WATER PUMPS

77

of the pistons vfbm the pump was working. In ^^tnIvius' descrip-
tion of the organ, where the pump is used to compress air, it is
stated that the pistons should be 'wrapped in sheep's hide with
the wool still on it', but there is no mention in Hero or '\^tniviu8
of a leather washer on the piston of the water pump, which is
surprising.

The inlet ducts were in the form of round holes in the bases of
the cylinders, over each of which (inside the cylinder) was a non-
return valve of a very simple type. It consisted of a disc, which,
when held against the inlet during the compression stroke, more or
less sealed it ofT. The disc was held loosely in podtion by four pins,
which passed through holes near its edge, so that it could release
the seal during intake, but could not become displaced, and would
dose up immediately under pressure. Hero desaibes another type
of valve which) he says, the Romans called assarium ('penny-valve'
— an odd chcuce of name, as it would bdong much more appro-
priately to the disc type. Was Hero's Latin a bit shaky?) This
consisted of two bronze plates, about Jin (2cm) square, one of
them having a round hole in its centre. This plate was sealed onto
the outlet duct, and the other plate was attached to the first by a
hinge nmning along the top edge (Fig. 21). The two adjacent

surfaces were carefully smoothed to make a good seal. Both this
type of valve and the disc type were gravity loaded, and could
only be mounted in one position — the disc horizontal and the
assarium vertical. Reliable spring loading would be very difficult
to achieve with the materials then available. The best that Hero
could find for the organ key-mechanism was a strip of animal
horn.

The outlet pipes come from the inward-facing sides of the
cylinders, near the base. Hero's design has a T-pieoe with a vertical

Fig. 21. Pump valves.

Copyrighted material

78 ENGINEERING IN THE ANCIENT WORLD

pipe coming up from the middle, with assariutn vsdvcs on the out-
sides of the cylinders, but because Vitruvius uses outlet valves of
the disc type, it is necessary for each of the outlet pipes in his plan
to curve upwards and enter, vertically, the base of a shallow bowl
[catinum). Hie valves are fixed on the ends of the pipes, and an
inverted funnel is then placed over the bowl, fhinly fisDcd to it,
with the outlet pipe rising upwards from its spout (Fig. 22).

Fig. 22. Vitruvius' outlet valve system.

For water to be drawn in, it was clearly necessary for the bases
of the cylinders to be kept under water. In the fire-engine this
would be done, as it was with the wartime stirrup pump, by tipping
buckets of water into the tank (see below), but the whole pump
can, if necessary, be submerged to any reasonable depth. In a well,
for instance, in which the water level rises and falls seasonsdly, the
pump could be fixed below the lowest normal level, and would
work effidendy whatever the depth at any given time. The only
variation would be in the degree of lift required, and hence in the
power needed to work the pump.

Parts of several pumps of this type have been found, one at
Trier in Germany, and the remains of another, now in the British
Museum, comprise most of the two cylinders, the pistons, the T-
piece and parts of the disc-valves. Parts of another much larger
pump were foimd in Silchester and are now to be seen in the

Copyrighted material

WATER PUMPS

79

Reading Museum.* The whole structure was enclosed in a wooden
block (probably a tree-stump), most of which is preserved — due,
no doubt, to the fact that it was under water — and the carved-out
portions where the cylinders, outlet pipes and valve chamber were
fixed can be clearly seen. The closeness with which they fit Vitni-
vius' description is quite remarkable. Two long lead tubes also
survive, one still in position in the block and another apparently
identical, with traces of solder on it. These were almost certainly
the cylinders. Assuming that the pistons (nothing remains of them)
were not more than about Gin (15cm) in length, the stroke might
have been in the region of 14in (about 35 cm). This, with a bore
in the teffxm of 2in (about 5cm), would give a cylinder capacity
of just over If the pump were 100% effident, it would deliver
nearly two pints (just over 1 1) per complete stroke (up-and-down),
but there are so many incalculable factors which might have re-
duced its efficiency (leakage in the valves and cylinders, friction,
etc.) that it is not worth trying to calculate how far it fell short of
100% efficiency. However, at a very rough guess, and taking a
rather pessimistic view, it might have delivered something like 3
gallons per minute. At a head of 16ft (4.9m), which is about the
average depth of the wells at Silchester, this would represent quite
an easy workload for an able-bodied man, which he could keep up
more or less indefinitely. It compares quite dosely with the sort of
amount one gets from a garden hose in an area of average-to-low
mains water pressure.

Hero's fire-engine, described in Book I of his PneumaHcay
Chapter 28, involves a two-cylinder force-pump. It has a water
tank, for which Hero gives no measurements, neither does he
actually say that it was mounted on wheels, though it almost cer-
tainly was. Presumably he thought that the outward appearance
was familiar to his readers, and concerned himself only with the
unseen working parts. The cylinders of the force-pump were fixed
to the base of the tank, and the pistons were worked by a rocker-
arm pivoted on a post in the centre. The outlet pipe rose vertically
£rom the T-piece, and was fitted with a device to enable the nozzle
to be tilted up or down, and swivelled round in any directuxi —
obviously a vital feature, in the absence of flexible hoses. The

♦See Frederick Davies, appendix in Archaeologia 55 (Part 1) 254—6.
G. C. Boon, Roman Silchester (Max Parrish, London (1957) 159-61) and
Tricrci Zeitschrift XXV, 109-21.

Copyrighted material

80 ENGINEERING IN THE ANCIENT WORLD

device has two joints of the kind which Hero calls 'sleeved'
(synesnurismenon) — i.e. with one pipe fitted inside anodier,
tightly enou^ to prevent leakage, but loosely enough to allow the
unfixed pipe to twist around. The elevation adjustment is made

by the arrangement shown in Fig. 23, where the joints at A and B

arc 'sleeved'. Hero then explains that the nozzle must be able to
turn horizontally, 'otherwise', he says, 'the whole apparatus will
have to be turned round, which is slow and pretty useless in an

emergency'. So the vertical outlet pipe from the pump is also
'sleeved', enabling the whole top part of the nozzle assembly to
turn round through 360°. There is still one further refinement. A
collar is fixed around the outer pipe, and L-shaped lugs are fixed
on the inner one to engage witii it (Fig. 23). 'This', says Hero,
'prevents the uf^er pipe from being bbwn ri^t off the apparatus
by the water pressure'. In view ol these eminently practical com-
ments (what could sound more like bitter experience tiian the
second ?) the suggestion that the whole machine was an unusable
armchair invention seems rather absurd ; though it becomes more
understandable, if not excusable, when one looks at the ridiculous
diagrams of it in a number of textbooks.

Remains of a force-pump with vertical outlet pipe and rotating
nozzle, which closely fits Hero's description, were found in 1889
in the Sotiel Goronado mine in Spain, some distance below ground.
In archaeological textbooks and articles the device is usually re-
ferred to as the 'Valverde Huelva pump^ and it is housed in the
National Archaeological Museum in Madrid. The nozzle can be

Collar-

Fig. 23. Hero's nozzle system.

WATER PUMPS

81

turned horizontally and verdcally , and the only difference between
this pump and Hero's design — a trivial one — lies in the outlet
valves. Hero uses the assarium type, with a verdcal hinged flap,
while the Valvcrdc Hudva pump has both oudet and inlet valves

mounted horizontally. They are of the disc type, and a little more

sophisticated than Hero's disc valves, in that each is enclosed in a
small cylinder. Each disc has a short vertical stem which rises
vertically from its centre, and the top of the cylinder has sl central

hole throt^ which the stem slides up and down — a valve^guide,
in fact Around this central hole are other larger ones through
which the water passes when the valve is open (Fig. 24). In order
to make the mounting of the outlet valve horizontal, the outlet
pipe from each cylinder has a right-angle bend, and leads into a
chamber which corresponds to the cntinnm in Vitruviiis' design.

More recently, four bronze pumps of a slightly different design
were found in the 'Dramont D' wreck — a Roman merchant ship
dating from about the middle of the first century a.d.* It is not
clear whether these pumps were part of the merchandize being
carried, or whether they were the ship's bilge-pumps. If so, they
were apparendy not in use when the ^p went down* They show

*See G. Rouanet) Etude de quatre pompes d eau Rmaines, in Gahien
d'Arcfa6ologie Subaquatique III (1974) 49-79.

Fig. 24. Valvade Hudva pump.

82 ENGINEERING IN THE ANCIENT WORLD

a curious feature, which recalls the bucket-wheels in Spanish
mines. The various parts of the four pumps have distinguishing
marks (dots, arrows, etc) which appear to be guide-marks for the
correct assembly of the parts. These marks are all the more impor-
tant because the pistons and cylinders are not oi exactly standard
bore, and therefore are not intcrchangeaUe. At ^at stage &ey
were put on, and for whose benefit, is a matter for argument.
Perhaps they were put on in the actual foundry where the bronze
casting was done, and used by the craftsmen who assembled the
parts (mainly by soldering), even though they were probably in
another part of the same factory.

These pumps have yet another different arrangement for the
valves. Each cylinder is mounted on a rectangular box, with the
inlet valve at one end, the outlet pipe at the other, and the outlet
valve on a partition between the cylinder wall and the outlet end
(Fig. 25). Both ends, and the partition, slope at an angle to the
vertical of about 15°~20°. The valves are of the assarium type,
with a flap wliich hangs from two hook-shaped brackets at the top.
As each valve is on the upward-facing side of the sloping 'bed*, it
naturally tends to fall shut unless forced open by the flow of water.

P P

Fig. 25. Pump found on Roman merchant ship.

Copyrighted material

WATER PUMPS

83

Accurate measurements of these pumps give a new insight into
the competence of andent metal-workers, and faring sharply into
question some of the traditional arguments to the effect that

inadequate technology limited the progress of applied science.
The meaiurenients of piston and cylinder in eacli pump are as
follows :

Piston Cylinder CSeanuioe

outside inside all-round

diam (mm) diam (mm) (mm)

Pump No. 1 43-5 44 0*25

No. 2 42-8 43-5 0*35

No. 3 42-9 43 1 0*1

No. 4 44 44-3 0-15

The inner surface of each cylinder and the outer of each fnston

has been highly polished, probably with some form of grinding
paste, and if it is coated with heavy grease makes a very good fit.
Tests suggest that the overall efficiency of one pump, including the
valves, was about 95 % . Its output, pumping at a rate of one stroke
(up-and-down) per second, was about 140 gall (630/) per hour.

Of cruder construction, but on a more impressive scale, is a
pumping system discovered at St Malo in Brittany in 1971.*
This included an eight-cylinder force-pump. The 'cylinder block'
was made from an oak beam, the cylinders themselves being drilled
out in a line along it. In the language of autCMnobile-makers, it is a
'strai^t eight'. No trace of the pistons survives, and of the valve
mechanism we have only a set of dght short wooden tubes, which
fitted into ducts at the bases of the cylinders. Nor is there any trace
of the mechanism for working the pistons, and it is impossible to
tell whether they all rose and fell together, or four-and-four
alternately. Even the purpose of this pump is obscure. It was found
in a square tank connected to a small square reservoir, down near
the sea shore (in fact, the site is now below sea level). Perhaps by
some quirk of nature there was a supply of fresh water there, or
perhaps the pump served to fill up salt-pans. We shall probably
never know.

*See R. Sanquer in Gallia, Tome 31 (1973) Fasc. 2, pp. 355-60.

Ccpyiiylited

4

Cranes and Hoists

We have vaiioiis sources of information on ancient cranes. Actual
remains of the machines are almost completely lacking, but there
is abundant evidence in the form of buildings erected by means of
cranes, some of which must have had a quite remarkable lifting
capacity. The architrave sections of the Parthenon, for example,
weigh something in the region of 9 tons each, and had to be raised
to a height of about 34ft (10.5 m) for positioning on the columns.
Most of the coliimns are built up of 11 drums, each weighing
about 8 tons, which had to be lowered accurately onto a central
spigot. This example, from the latter half of the fifth century b.g.,
is chosen as the best known, though it is neither the oldest nor the
biggest. Apart from such evidence, we have an extended Latin
account of two types of crane (Vitruvius X, 2, 1-10) and a de-
tailed relief sculpture, probably of the early second century a.d.
It occupied a panel in the family tomb of the Haterii, one of whom
was apparently a building contractor."^ This sculpture has a
number of interesting features. The car\ing was done by a mason
who was familiar with the crane itself and had an eye for detail.
For instance, two men arc shown tying a rope around the top of
the jib, and the carving is so accurate that it is possible to identify
the knot as a reef-knot. As an artist, he leaves much to be desired.
His perspective is hopelessly bad, and the buildings are lumped
together in an almost surrealisdc way. But this does not detract
from the value of his evidence on technical matters.

Vitruvius' description of the first type of crane is as follows.
Two beams are required for the jib, their thickness depending on
the maximum probable load. They are fixed together at the top
with an iron bracket, and separated at the base, like an inverted V

^Illustrated m many places, e.g. G. E. R. Lloyd Gnek Science after

Aristotle (in this series) Chatto & Windus, London 1973, fig. 24, p. 110.
The best photographs are in Atti della Accademia Nazioaale dei Lincei,
series VUI vol 13 (1968), figs. 15 and 17.

Copyrighted material

CRANES AND HOISTS

85

(hence the modem name 'shear-legs'). Ropes are attached to the
head of this jib, and arranged 'all round' to keep it steady, A
pulley block is suspended from the top with two wheels, one above
the other ; the hoisting rope passes over the higher of these, down
and around the (single) wheel of the lower block, which is attached
to the load, up again and around the lower pulley of the upper
block, and down again to an eye on the lower block (Fig. 26).

Fig. 26

The other end is brought down between the legs of the jib to a
windlass turned by handspikes (four, all at right-angles). From the
lower pulley block iron forceps are suspended, the teeth of which
fit into holes in the blocks of stone to be lifted. This arrangement
is called a *triple-haur (in Greek, trispaston). If the upper block
has three pulleys and the lower block two, it is caUed a 'quintuple-
haul' (pentaspaston).

Copyriyhica niaiciial

86

ENGINEERING IN THE ANCIENT WORLD

If the crane is required to lift bigger loads, the beams for the

porticmately strengthened. Vitnivius then describes a technique
for raising the jib from the ground, using a second windlass with
a stay-rope passing over a pulley at the top of the jib, over the
'shotdder-blades' {scapulae, presumably raised projections on the

back of the jib, to reduce the initial mechanical disadvantage), and
via a double-pulley system to an anchor-pcnnt some distance away

or, if there is none available, to a set of inclined piles driven into
the ground. Once the jib is raised, adjustable stay-ropes on all
sides can be attached, each with its own anchor-point. If the wind-
lass and quintuple-pulley system are not powerful enough to raise
the load, an additional reduction gear is used. In place of the ^vind-
lass there is a shaft passing through a drum, fixed at its centre.
The hoisting cable is doubled, its two ends winding onto the shaft
to the right and left of the drum, and the pulley blocks have pairs
of wheels side by side instead of single wheels. A secondary rope

Fig. 27

Copyrighted material

CRANES AND HOISTS

87

b wound round the circumference of the drum, and taken to a
windlass, probably mounted below on the jib (Fig. 27). Alter-
natively, the drum can be made large enough to act as a tread-
mill, worked by men from inside. If so, it would be mounted at one
side of the jib, on the end of its shaft (Fig. 28).

Fig. 28

This probably represents a fairly advanced stage in the develop-
ment of crane design. The later centuries of the Roman Empire
may have impro\'ed on the details, but the fact that an illustration
of a century later fits this description almost exacdy suggests that
the design was standardized. It had certain merits and limitations.

The forked jib, unless some pivot mechanism was fitted on the
base (and Vitruvius does not mention any), could only swing for-
ward and backward along a straight line over a limited distance.
If the jib is raised too near the vertical, the load strikes against it,
and may damage it or the lifting tadde. As the jib is lowered
forward from the vertical, the stress on the rear stay-ropes increases,

Copyrighted material

88 ENGINEERING IN THE ANCIENT WORLD

until at an angle of about 30° from the vertical (depending on the
distance of the anchor-point) it becomes equal to the load being
rsused, and beyond that point it exceeds the load. So the forward
tilt of the jib, and hence the range over which the crane can move
its load, arc limited by the strength of the stay-ropes and of the
mechanism for adjusting them. This is why Vitruvius says that
the jib must be longer for a gieater load, since it can then move
its load forward over the same distance with less deviation from
the vertical.

The second design de^c rihcd by Vitruvius has a single-beam jib,
and he points out that it can be swung sideways as well as
forwards. However, its handling capacity was probably a good
deal less. At the opposite extreme, a tripod derrick, mendoned by
Hero in the Mechanica (HI, 4) would have no mobility at all.

The stay-ropes at each side are required to prevent toppling,
especially when the load is not initially central and swings about
when lifted from the ground, but it is absolutely essential to have
another running forward from the jib-head, as is shown in the
relief ilkistration. The danger is that as the jib is tilted upwards
towards the vertical, the load on the rear stay-ropes diminishes
quite suddenly to zero. Without the forward stay-ropes there is
nothing to prevent an over-enthusiastic heave from bringing the
jib and the load right over backwards and down onto the crew.

Hie disadvantages of the multiple-pulley system lie in the loss
of energy through friction in the pulleys and creep in the rope. The
advantages are that it not only provides a reduction ratio of 5:1
(with a pentaspaston), but also distributes the weight of the load
evenly between five sections of the lifting cable, so that a rope with
a capacity of (say) J ton can be used to lift up to 2^ tons, or 5 tons
if doubled. The gearbox system described by Hero of Alexandria*
has a reduction ratio of 200:1 — or more, if a worm gear is used
— and could be used instead, but the load is attached directly to
the final shaft, and a single thickness of extremely strong rope
would have to bear the entire load. It is difficult to support the
view of some scholars that this gearbox was merely a pipe-dream
— pipe-dreamers do not allow for friction in the gears, as Hero
does. However, it may well be that for practical reasons the older
and less sophisticated pulley system was preferred. The rope-
driven reduction gear is simple, and was presumably as cheap to

♦sec pp. 206-7.

CRANES AND HOISTS

89

make as a gear system, but it had its problems. To raise the load
10ft (3 m} on a five-puUey system, 50ft (15 m) of the hdsdng cable
must be wound up, and supposing the windlass shaft to be about
Gin (15cm) in diameter, this would require about 32 revolutions

of the shaft, ending up with a lot of rope wrapped around it. If
the drum were 3 ft (90 cm) in diameter giving a reduction ratio of
6:1, 300ft (90m) of the secondary rope would have to be unwound
from it — 32 turns again, though of thinner rope in this case. If
the second windlass were the same size as the shaft, it would have
to be turned nearly 200 times, and the bulk of rope around it
would become very large. Perhaps the rope was wrapped a few
times around the windlass shaft, held under tension to avoid
slipping, and stowed in a heap instead of being wound onto the
shaft.

If the handspikes on the final windlass were 3ft (1 m) long the
total reduction ratio would be 180:1. Even allowing for c(»isider-
able loss of energy in the pulleys and the rope — say 30% — one

able-bodied workman could probably hoist at least 2 tons with it.
It would take some time, perhaps half an hour to raise it 10ft, but
this could be speeded up by extra men, up to four if there were
handspikes at each end of the windlass.

It seems highly probable that the weak point of the whole
apparatus was the attachment between the lower pulley block and
the load. The 'iron forceps* {ferrei forfices) are not visible in the
illustration. But we have a brief reference to them in Hero's
Meckanica III, 7. The details are by no means dear, but
apparently the device was called a 'crab', and had three or
four 'jaws' (three for column drums, and four for rectangular
stones) with bent ends. These were either pushed under the
bottom edge of the stone to be lifted, or into holes specially cut in
the sides. Then wooden cross-pieces were fixed on (perhaps tied
on with ropes ?) to prevent the jaws coming apart and allowing the
stone to slip through. The holes in the stone blocks must have been
difficult to conceal unless the stone was deeply carved after being
laid in place, which seems unlikely. If they were cut in the ends of
the block, they could be concealed by adjoining blocks in ashlar
masoruy, but they would have to be laid to one side of the final
position for the jaws to be removed, and then shifted along by
other means. Rollers were apparently used for this on some
occasions. Moreover, the danger of a block splitting in mid-air

Copyrighted material

90 ENGINEERING IN THE ANCIENT WORLD

and falling from the forceps must have been ever-present. Hero
points this out as one of the hazards for the crane crew.

Some other methods of attaching the load are known from
ancient evidence. One was to leave projecting lugjs on the sides of
stone blocks or column drums, whidi could be trimmed off when
the stone was laid in place. On some column drums the lugs appear
large enough for ropes to be tied around. The fact that they are
below the centre of gra\aty docs not constitute a problem. Once
the hoisting ropes were looped over the lugs, two other ropes could
be lied around the circumference of the drum, one just above the
lugs to keep the hoisting-ropes from slipping off them, and the
other near the top of the drum to prevent it from tipping over
when lifted. In fact, it would be safer to lift it in this way than to
have the lugs higher up, since the weight of the portion of the
drum below the lugs would put an eictension stress on the stone,
and might crack it if there were any flaws in the centre. Alterna-
tively, the lugs might have provided a grip for forceps.

Another method of attaching the load was to make a sling of
ropes passing under the stone. The problem here, of course, is to
get the sling away when the stone is laid in place. When working
with small blocks as part of an ashlar wall, crowbars could be

used to jack up one end while the ropes were imtied, and marks

(known as pry-holes) have been found on a number of blocks
which suggest that tliis has been done. But two other methods
were devised in antiquity for dealing with the problem. One was
to cut a deep I T-shaped groove in each end of a block (Fig. 29).
The ropes passed through these grooves and were prevented from
slipping out by a slight flange on the upper wail of the groove. The

Method ofaUackhig rope
.leaving end flush

Fig. 29

Copyrighted material

CRANES AND HOISTS

91

blocks cotild be laid flush together, and the rqpes untied and re-
moved. This ingenious method has two limitations. As Greek
ashlar stonework was not normally cemented, it was possible for
water to gather in the groove which, in exceptionally cold condi-
tions, might freeze, exjjand and so displace the stones. Secondly,
the method could only be used on adjacent surfaces which were
eventually to be concealed, and was not suitable for corner-pieces,
the ends of courses, or (as a rule) exposed top courses.

Fig. 30

Yet another method was to use what is known as a *lewis bolf,
or a 'lewis iron', or just a lewis*. We have no actual remains of
these from andent times, but a brief reference in the next chapter

of Hero's Mechanica (III, 8), taken with the archaeological evi-
dence, makes it clear that they were essentially the same as modern
ones. A rectangular hole has to be cut into the top surface of the
block to be lifted, of the order of 1 ft long by Sin wide (30cm X
12 cm). As it goes down through the stone, it gets wider and wider
(Fig. 30), its sides sloping out at an angle of about 4-5°, in a
dovetail shape. The bolt itself is in the form of three iron bars,
two of them bent over at their bottom ends *in the form of the
Greek letter J^, as Hero puts it. These two were inserted into the
hole, the bent ends f adng outwards, and the third (flat) bar was
then pushed between them, holding them apart and preventing
them from slipping out of the socket To hold them all together
as one piece, holes were drilled through thdr top ends and a pin

uopyiiyhiea maiciial

92 ENGINEERING IN THE ANCIENT WORLD

pushed through. The most likely method of attaching the hoisting
cable was by a stirrup with holes in its ends, through which this
pin also passed (Fig. 30). Hero notes that the quality of iron used
for the bars must be carefully checked. If it is too hard (forged for
too long with a heavy hammer?) it will crack, and if too soft
(annealed?) it will bend. In either event, the failure is most likely
to occur at the bends in the outer bars. A happy medium between
too hard and too soft must be found.

The advantages of the lewis iron are obvious. It can be used to
place a block on a flat bed, without putting anything underneath
it or having to move it into its final position, but its disadvantages
are serious. Cutting the holes is a long and tedious business, re-
quiring some skill. When under load, the lewis iron acts as a wedge,
and the risk of splitting the block (already weakened by the cutting
of the sockets) is considerable. This is why at least two bolts were
normally used in all but the smallest blocks, to distribute the stress.
Finally, when the job is done, the block is IdPt with a large unsightly
hole, which has to be covered over or filled in by some means. In
view of all this, it is not surprising that evidence for the use of the
lewis iron is not copious, and is ahiiost entirely confined to build-
ings which have been dismantled and rebuilt after damage
(e.^. by earthquake). In such situations there is no alternative
method, since the lugs have been removed.

The heaviest blocks used in any building are usually the archi-
tra\'es and lintels, which can be hoisted into place using a rope
sling without raising a problem over its removal. They do, how-
ever, require very accurate placing, and this may be difficult if the
weight is very great. An ingenious method was devised by Gherd-
phron, the architect of the temple of Artemis (Diana) at Ephesus
in the sixth century B.C. He had a mound of sandbags erected
between the pillars, rising a little above the capitals, and had the
architrave 'raised onto it'. Some scholars have taken this to mean
that the sandbags formed an inclined plane, and the architrave
was hauled up it. It seems rather incredible, however, that a pile
of sandbans, ils width limited to the space between columns, could
have been stable enough to stand this. Pliny's text at Nat. Hist.
36, 97fT. (the source of our information) is doubtful, and could
mean *a gende slope' or 'a soft support'. It is much more likely
that the architrave was hoisted by crane and placed roughly in
portion, when it could then be adjusted very finely by letting sand

CRANES AND HOISTS

93

out of the lower bags, thus causing the pile to settle very gradually
in the required direction.

The design of the crane illustrated on the monument of the
Haterii is almost exactly in accordance with Vitruvius' description.
The crane has a quintuple pulley system on the hoisting cable, and
forward and rear stay-ropes, each with a triple pulley. The pulleys
are each constructed from two iron plates, held together by the
pulley axles and, in the main hoisting pulley, by a tie-bar at the
top, around which the strop passes. The very scanty remains of
ancient pulleys whicli do survive are made of wood.* The machine
is not actually in use, but being prepared for use, and the main
task of the two men 'up aloft' would normally be to over-haul the
pulleys. They are shown tying a rope around what looks like a
wicker basket placed updde-down over the tip of the jib, and there
is a dump of foliage behind, but it is impossible to say whether it
is being tied onto the crane (in some kind of *topping-out' cere-
mony ?) or whether it is merely for decoration in the comer of the
sculpture. The basket may have been a weather-protection for the
wooden jib or, more probably, a buffer to avoid damage to stone
carving in case the jib accidentally struck a completed part of the
building. The workmen are barefoot ^ — a sensible precaution for
climbing around on ropes and pulleys — and one of them appears
to be wearing a leather cap — a forerunner, perhaps, of the crash
helmet.

Five men are shown climbing about on the wheel. It would be
reasonable to suppose that in actual use two men would provide
the main lifting power (perhaps three for an extra heavy load),
and one would stay at or near tiie bottom, in communication with
a foreman up on the wall or building who could see the exact
position of the load. The third man below would make the fine
adjustments of height by his own w'eight, in close co-ordination
with the other two. To do this properly would require a lot of skill
and extended practice — nothing could be easier than to bring
about a very serious accident. No matter how plentiful the supply
of slaves may have been, a trained crew must always have been
at a premium, and very difficult to replace.

The second type of crane described by Vitruvius (for which
we have no ancient illustration) had a single-beam jib instead of

*J. W. Shaw, A DoubU-^moed PulUjf Bhek Jhm Cenehrtae, Hesperia
XXXVI/4 (1967) 389-401.

94 ENGINEERING IN THE ANCIENT WORLD

shear-legs. It was held upright, or tilted forwards or sideways as
required, by four stay-ropes, and three hoisdng cables, each with
its own quintuple puUey system, passed over pulleys at the base of
the jib and were hauled b) three teams of men. By dispensmg with
the windlass the speed of operation was greatly increased, but it
would clearly require, as Vitruvius says, a highly trained and skilled
crew to work it successfully.

Fig. 31

At the very end of the chapter, Vitruvius casually remarks that
other types of crane are used for various purposes, particularly on
docks for loading and unloading ships, 'some of them upright
[erectae) and others horizontal {planae), with revolving carchesiiL*
This cryptic sentence requires some examination.

The word carchesion can bear a number of meanings. In the

Copyrighted material

CRANES AND HOISTS

95

coDtext of ship's rigging it seems to mean something like a crow's
nest, or the block in which (or to which) the pulleys for the halyards
were fixed. From its shape it also gave its name to a particular
kind of drinking-cup. At first sight, therefore, one might be tempted

to assume that Vitruvius is talking about a gaff, such as can be
seen to this day loading and unloading barges and small \ csscls.
This consists of a spar, with a fork at its base, which rests on a
wooden collar high up on the ship's mast, with the tip of the spar
supported by a rope running to it from the masthead. The hoisting
cahie passes over a pulley suqiended from the tip of the spar, and
another rope or ropes are used to swing the spar round sideways,
over the ship's hold or over the quay as required (Fig. 3 1).

There are, however, a number of considerations which cast
doubt on this. The most telling piece of evidence comes £rom a
passage in the historian PolyHus describii^ the siege of Syracuse
by the Romans in 212 b.g. (He was writing in the second century
B.C., perhaps 60-70 years after the events.) He describes two ways
in which cranes designed by Archimedes were used as weapons of
war. One was to drop heavy stones or lead weights on approach-
ing ships (the sea came right up to the battlements), and he says
they were 'swung around as required on (or by means of) a car"
chesion\ In the absence of any luffing device* (and there is no
evidence for it) it must have be^ possible to swing the jib forward
and backwards, to get the missile directly over the enemy ship. If
it could swing sideways as well, this suggests that carehesion bears
the same meaning that it has in relation to a catapult — namely, a
swivel mounting with bearings at each side (Fig. 32). This is con-
firmed by the next paragraph, in which Polybius describes a
different use of cranes. He tells how landing-j)artics, on board
ships fitted with arrow-proof protective screens, were pre\ciited
from approaching. 1 hey were driven back from the prow of the
ship by missiles (stone shot, he says) and then an iron grappling-
hook was lowered on a chain. The man controlling the jib (literally,
'steering', the word normally used of a ship's helmsman) tried to
hook it on the ship's prow. When this was done, he *drew down'
(or perhs^ 'pressed down') the opposite end of the jib, which was

*This enables the point of suspension for the load to be moved out from
or in towards the base of the crane. In the modern tower-crane to be seen
on many building sites it takcf the fiurm of a smaU trolley which runs back
and forth on rails along horizontal boom.

Copyrighted material

96 ENGINEERING IN THE ANCIENT WORLD

inside the wall. (The word used is ptema *heel', used of the butt
ends' of catapfuh anns^ see p. 1 1 7.) Tliis clearly implies that the jib
of the crane pivoted, not at its base as the shear-legps did, but
(probably) not far from its mid-point. This lowering of the butt
end', says Polybius, lifted the ship's prow out of the water, and
stood it up vertically on its stem'. Next, the crane operator 'fastened
the machine to make it immovable, and then, by some sort of
release mechanism, cast off the grappling-hook and chain. The
ships then either capsized, or listed badly, or (Polybius could not
resist the temptations of rhetoric) 'became filled with confusion and
much sea-water.'

Fig. 32

What are we to make of this account? It could, of course, be
complete ficdon, but the chances are that it contains a nucleus of
truth which has gathered an accredon of fantasy in the telling. In

a later account, the Roman commander Marccllus is reported as
saying that his ships had been 'sloshed about like wine-ladles'.
This, even though it comes from Plutarch {Marcelhis, 17),
might be authentic ; Marcelhis was trying to save face, and justify
his action in abandoning a direct attack and turning instead to a
prolonged blockade.

Polybius gives the impression that the ships were hoisted right
out of the water, but this is most unlikely and in any case quite

Copyrighted material

CRANES AND HOISTS

97

uimeccasary. If the landing-party had been forced away from the
bows, the ship would be down by the stem to begin with, and
then it would only be necessary to hdst the bows far enough, and
give the hull sufficient list, for water to be shipped over the stem

— the laws of hydrostatics (and who should know better than
Archimedes?) would do the rest, without any additional pull from
the crane. I he ship would end up 'standing vertically on its stern',
and although most of the crew would by then have jumped or
fallen overboard, the buoyancy of the wooden hull would probably
find equilibrium with not more than a few feet of the bow out of
the water. If the crane had lifted it to a higher level than this, the
release of the grappling-hook and chain would cause the hull to
plunge quite violently, and even if it righted itself afterwards, the
crew would be in the water, the ship awash and useless.

Cdrchesion

Fig. 33

We are left, then, with an impression of a crane mounted on a
tower slightly higher than the fortifications, with a boom on a
swivel mounting which could tip up and down, and which, if not
pivoted at the centre, might have been balanced with some sort of
counter-weight. A type of scaling-ladder (sambuca) described by
Biton was constructed in precisely this way.* The hoisting arrange-
ments can only be guessed at. One suggestion, which depends
heavily on Polybius' expression' lowered the opposite end, which

•See J. G. Landels, Shipshape and Sambuca-fashion, J.H.S. LXXXVI
(1966), p. 72.

98 ENGINEERING IN THE ANCIENT WORLD

was inside the wall', is that a block-and-tackle was suspended
between the far end of the boom and an anchor-point at llie base
of the tower, and a tug-of-war team pulled aa the cable (Hg. 33).
Provided the anchor-point was near enough to the axis on which
the boom turned, it could be swung around over a limited arc
while under load. A similar arrangement (apart from the tug-of-
war team ! ) can be seen on some small mobile cranes used by
building contractoi's. Such a design would fit Vitruvius' description
('horizontal, with revolving carchesion\ and would be well suited
to the dockside work to which he refers.

Copyrighted material

5

Catapults

A NUMBER of ancient treatises, and a few surviving iliustradons,
together present a coherent picture of the deagns of various
weapons of this type, and of the ways in which those deagns were
gradually developed and improved. The most important written
sources are the works of Hero of Alexandria and Fhilo of
Byzantium.

Tlie precursor of all types of military catapult was the bow,
known in the eastern Mediterranean from remote antiquity.
Though very simple, the bow is a highly efficient device which
can store the energy imparted to it gradually by the action of
drawing, and release it very rapidly. Moreover, the geometry of
the string is such that the missile is given rapid acceleration and
(to borrow a term from other weapons) a high muzzle velocity.
Hie amount of energy which can be stored in a bow is determined
by two factors — the stifiEness of the bow (i.e. the force required to
bend it) and the length of draw. Both of these are in turn subject
to human limitations. The first depends on the strength of the
bowman's chest, arm and shoulder muscles — a bow requiring a
draw force of 1001b (45 kg) is considered to be near the Ihnits of a
normal-sized man's capability. It also depends on the ability of
the fingers of the right hand to control the string, to hold it during
aiming and release it at the right moment. The second factor is
limited by the length of the bowman's arms, from the left hand
fully extended to the right hand, drawn back beside the right
shoulder. It may be said with some truth that the history of military
catapults is the history of the engineers* attempts to overcome these
limitations.

Their earliest attempt (according to Hero, and there is no reason
to think him mistaken) was something like a crossbow^ and went
by the homely name of ^beUy-shooter* (in Greek, gastrt^hitis) loc
reasons which will become dear. The date of its inventioa is un-
certain, but such evidence as we have suggests that it was not

Copyrighted material

100 ENGINEERING IN THE ANCIENT WORLD

before the fourth century b.c. The weapon was almost certainly
not used in the Peloponnesian War between Athens and Sparta
(431--4r04 B.a)» since Thucydides^ the contemporary historian of
that war, who shows obvious interest in si^|!C-engmes and unusual

weapons, makes no mention of it.

The basis of this weapon was a bow, probably not much longer,
if at all, than an ordinary hand-bow, but rather stifTer — too stiff,
in fact, to be drawn by an archer's hand. This brings us right
away to a problem which keeps recurring in the study of ancient
technology. The designer of the catapult, since he was expounding
something new and complicated, described it in detail and pro-
vided diagrams, but the bowmaker was a craftsman. He made his
bows very much as his father had before him, any improvements
being strictly empirical, and he did not write books about it — his
apprentices learned their craft by word of mouth and practical
demonstration. The engineer is content to specify in his plans a
bow of a certain size, without giving any details of precisely
how it was made, or even of the materials used. Some social
historians interpret this as evidence of class-consciousness — the
engineer thinking of himself as an intellectual and of the bow-
maker as an artisan. But there are other possible esqilanations,
one of them being that Hero was probably justified in assuming
that his readers knew exactly what a bow looked like, and how
it was made. A similar problem arises later over the making of
sinew-rope.

It is, however, generally agreed that the Cireek bows from which
catapults were first developed were composite, and there is, at
least, some evidence that horn was used in their construction. If
so, they were perhaps made in three layers, with a central lath of
hard wood, which served as a core, sandwiched between the other
two, one of horn on the inade of the arc, which was compressed as
the bow was bent, and the other of sinew on the outside, vAddi
was stretched. Primitive communities in more recent times have
used this type of bow with the sinew either anchored at the ends
of the core, or glued onto it. In the absence of any evidence, we
can only guess at the Greek methods. The famous passage in Book
XXI of the Odyssey (388-430) suggests that Odysseus' bow had
a layer of horn, and this must refer to the seventh century b.c. at
the very latest. Phiio also (Chapter 71) speaks of *horn and some

Copyrighted material

CATAPULTS

101

kinds of wood' as the materials from which bows are made. It is
curious that he does not mention sinew.

The designer of the 'belly-shooter' was faced with three prob-
lems; how to draw the powerful bow, how to hold it drawn while

the arrow was aimed, and how to release the bowstring at the
right moment. For an account of his solutions, one might do worse
than quote Hero in a literal translation, with a reconstruction of
his diagram which, unfortunately, has not survived.^ (Fig. 34.)

Fig. 34. The bow catapult (gasfyraphitis, belly shooter).

liet the bow in question be ABGD, with its curved ends AB
and CD too powerful to be drawn back by human hand. AD is

the bowstring. Fixed to the bow at the centre of its concave curve

is a batten EFGH, which has on its top surface a dovetail gioove
K L. Fitted in this groove is a male (dovetail) of the same length,
and lixed to its upper surface is another batten of the same length

♦The passage comes from his Belopoeica pp. 75-81 Weschcr, Alarsden
71 T. pp. 20-22. This translation follo^rs Marsden's text, with a few
trivial omissioiis and a few explanatory comments. English letters are
substituted for the Greek ones on Hero*s diagram.

Cc, , y od material

102 ENGINEERING IN THE ANCIENT WORLD

and breadth as £FGH; along the centre of its upper surface is a
semi-circular groove of the same length as the dovetail KL, in
which the mis^ is placed for firing. On the remaining part of the
top of this batten (i.e. between the tail of the missile and the rear
end) are two upright iron stanchions, naikd on and joined to each
other at thdr bases, quite dose together. Between llicm is placed
an iron claw, bent downwards at the end towards K (the front),
and its bent extremity should be divided to form two prongs, as
in the so-called skendylion. (TTiis was a shipwright's tool, of which
little is known ; but it was probably something like a tack-lifter or
claw-ended jemmy.) The space between the prongs should be wide
and deep enough to take the shaft of the "riisRilf!. A round pin is
inserted centrally through the stanchions and the daw. This claw
is marked MNO in the diagram, the two prongs at M and the
pin I. Bdow the opposite end NO is placed an iron lever PQ»
pivoting on a pin at P, which is fixed vertically on the top surface
of the upper batten. When the lever PQ is puiahed underneath, it
wedges the daw so that it cannot tip up, but when we take hold of
the end Q and pull on the lever in the direction of NO, then the
front end of the claw — the section MN — can tip upwards. Fixed
to the batten EFGH (at its rear end) is a bent bar RSTUV,
concave at S and convex at U and V.'

Hero then gives a number of standard terms which 'they' (docs
he mean the designers or the soldiers?) use for the various parts,
such as didstra ('slider' or 'push-through') for the upper batten,
and syrinx (*trough*) for the lower. He goes on :

'The machine was constructed as described above. When they
wanted to draw the bow, they moved the didstra forwards towards
K, until the claw tipped up and rode over the bowstring which,
when at rest, was just above the didstra. Then they tipped the daw
forwards, and pushed the lever under its rear end, so that it could
not tip up again. Next, they propped the front end of the didstra,
which was then pushed out at the far end, against a wall or against
the ground and, grasping the cuds of the bar RSTUV, they
pushed their stomachs against the hollow part between U and V
and, using the force of their whole bodies, pushed back the didstra,
thus pulling back the bowstring and bending the arms of the bow.
When they thought the tension was sufficient, they placed the
missile in the groove, and released the claw by jerking the lever
away from under it, and so the missUe was despatched.'

Copyrighted material

CATAPULTS

103

Here we have solutions to two problems — how to draw the
powerful bow, by using the whole body instead of only the anus,
and how to control and release the bowstring by means of a metal
device instead of the archer's fin^^ers. But the third remains, and

it may appear to the reader that Hero is guilty of a disastrous
omission. In fact, however, he is simply following his usual practice
of breaking up the description into clearly defined sections, and
avoiding anticipation of what is to come in the next paragraph.
He goes on to say that something must be done to prevent the
bowstring from pulling the diostra forwards again when the
pressure is taken off the stomach-rest. It must be held back until
the missile has been placed in the groove and aimed at the given
tai:get. This was done by fixing strips (probably of metal, thou^
he does not say so) along the sides of the ^ough' EFGH, with saw-
tooth ratchets cut in thdr top edges, and fixing pawls on the ades
of the diostra, which rode over the teeth at an oblique angle as the
diostra was pushed back. At the appropriate length of draw
each of the pawls was propped against the nearest tooth on the
ratchet. The fact that the Greek word used for pawl was korax,
meaning 'raven', might suggest that the pawls were hooked, but
this is not necessarily to be assumed.

Hero ends by saying that they called the whole weapon a 'belly-
shooter', because the stomach was used to draw back the bow-
string. This last remark should apparendy be taken to mean that
the gastraphitis was not so called because it was fired from the
stomach. It would in any case have been extremely difiicult to aim
with any accuracy in that position. The only World War TI
weapons normally fired from the hip were the Thompson and
Sten sub-machine guns, and then only in bursts and at very short
range. The gastraphetes fired single shots rather slowly and would
not be used at close range except in dire emergency. Incidentally,
the method of cocking it may remind some old soldiers of painful
experiences with the P.I.A.T.

This design did not merely solve the problems of holding and
triggering; it also — potentially at least — overcame the restriction
on the distance the bowstring could be drawn back, and hence cm
the length of the bow. IVovided it was strong enough not to bend
or break under the thrust, the diostra and the groove in v^iich it
slid could be made as long as desired. But the problem of bow-
stiffness had been only partially solved. Though capable of a

Copyrighted material

104 ENGINEERING IN THE ANCIENT WORLD

stronger thrust than the arms, the stomach — even that of a hefty
artilleryman — has its limitations, and the next stq> was to devise
another method of drawing back the diostra, and bending a bigger
and better bow. It was quite amply done ; instead of being pushed,
the diostra was pulled from the rear end by a small windlass
and rope. This gave a mechanical advantage equal to the ratio
between the radius of the windlass shaft and the length of the
handspike. Supposing the shaft to be 4 in (10cm) in diameter, and
the handspike 20 in (50cm) long, the ratio would then be 10:1. If
(at a reasonable guess) the artilleryman had been called upon to
put a thrust of 2001b (90 on the stomach-rest, imagine his
relief when, allowing for some friction, a puU of only 221b (10kg)
on the end of the haoidspike would draw a bow of die same stiff-
ness. Of course, all advantages must be paid for, and the price in
this case was speed. Three or four complete turns of a capstan
would take longer than a grunt and one hefty shove. Here is a
striking example of a situation in which the crank would have
been very useful, but Hero calls the handle skytale^ which can only
mean a rod or spike. Slowness, however, was almost cci Laiiily
compensated by an ad\'antagc which the new design ofTcicd at
the same time. We must assume that the gastraphete s w^zis usually
fired over a wall, resting on the top for aiming and pushed against
the base for loading. pA cn if it had some sort of stand or aiming-
rest, it would have to be lifted down between each shot. But once
the windlass was fitted, it could be kept in the firing position all
the time. Incidentally, the weapon was still called gasliraphHis,
though no longer in fact a Hbelly-shooter', just as rifle shooting in
World War I was still called 'musketry'.

The engineers had now reached a stage at which the best
general layout had been found, and the only way in which the
weapon could be improved was simply by increasing the size and
stiflness of the bow, and the strength of the other components
proportionately. This, however, did not increase the muzzle velo-
city very much, even when a light missile was used, since the
inertia of a big, heavy bow would prevent the rapid release of its
energy. What it did make possible was the use of much heavier
missiles over a comparable range — the heavy bolt and the ball.
In the ancient sources there are descriptions of bolts about 6ft Gin
(2m) long and about l^in (3.6cm} diaoneter. Ftesumably they had
heavy, sharp metal points, and would be fired on a low trajectory.

Copy I iyl uuU I Mdj-I lal

CATAPULTS

105

where lliey could be aimed with some accuracy. They would
certainly penetrate any normal body armour, and in many cases
the armour shielding (such as layers of hide) used on siege engines.
Even a wall or wooden stockade, a perfecdy adequate defence

against ordinary archery, might be quite useless against them. The
other form of missile was a round, or near-round, stone shot. These
have been found on ancient sites which are known to ha\ e been
under siege — or prepared for it — at the time when these machines
were in use, and many of them fall within the weight range of
1 1-55 lb (5-25 kg). They would almost certainly be hred, howitzer-
style, on a high trajectory to give maximum range and, falling
from a height of perhaps 150ft (45 m), would make a very powerfid
impact. Nothing less than a heavy, strong stone structure would
zSord any shelter, and in an age innocent ci explosives the destruc-
tive effect must have been immensely impressive. On the other
hand, the aim cannot have been very accurate.

The catapults for tiirowing stone shot were essentially the same
as the arrow-shooters, with two slight modifications. On the smaller
machines the bowstring was made flat, in the form of a strap, with
a loop on its rear side in which the trigger claw engaged. Accord-
ing to Hero, this loop was ^voven into the fabric of the bowstring.
Its height above the didstra was adjusted so that it would thrust
on the stone shot centrally, and not ride over the top or sUp under-
neath. On the bigger machines the bowstring was made in two
halves, with a leather or hair-cloth sling at the centre, in which
the shot was placed before firing, and the rii^ on the back of the
sling was metal. Then, as the bows got bigger and bigger, the
windlass was fitted with handles at each end, so that two men
could turn it, and was given additional mechanical advantage by
the use of a block and tackle. With a ^quintuple-haul' block (see
Chapter 4), two men could probably have used a bow with a draw
force of 2—3 tons. It may well be asked — did the Greeks of the
Hellenistic age ever succeed in making bows on that scale? Here
we have a few scraps of evidence from Biton. According to him,
one Charon of Magnesia (date unknown, but earher than second
century b.c.) designed a stone-thrower with a didstra about 1ft
(30 cm) wide. From the size for the sling, Dr Marsden estimated
that the stone shot might have been about 4iin (11.5 cm) in
diameter, and 51b (2.3 kg) in weight. The bow on another machine
of comparable size, again according to Biton, was about 9ft

Copyrighted material

106 ENGINEERING IN THE ANCIENT WORLD

(2.75 m) long and 3^ in (nearly 9cm) in diameter at its thickest
point. This gives a very rough idea ci the general scale and pro-
portions of these machines.

The question arises whether the bigger bows were oompodte,
with or without a sinew layer. Two scientific facts are relevant.
The energy-storing capacity of horn is greater than that of wood,
and the capacity of sinew is many times greater, so that a simple
wooden bow would have been very inferior in performance. More-
over, the energy-storing properties of wood are severely affected
by temperature. Above about 25°C (77°F) performance falls of!"
sharply, and such temperatures must often have been reached in
the Mediterranean area during the summer campaigning season.
The traditional English longbow, made of yew wood, is suited
only to the English — or northern French — climate.

In the next chapter, Biton describes an even bigger stone-
thrower, designed fay Isidorus of Abydos (date likewise unknown).
In proportion to the measurements given, this should have had a
bow about 15ft (4.6m) long and 1ft (30cm) in diameter, and fired
a shot about 9in (23 cm) in diameter, weighing something in the
region of 401b (18kg). Strong evidence that these dimensions are
in the realm of fact, and not just an engineer's line of sales-talk,
can be seen in the discovery of an ammunition store in Pergamon
by the German excavators,* which contained more than 250
stone shot of about this size. One other, really monster catapult
deserves at least a mention. According to a near-contemporary
account, it was fitted amidships on a very big armed merchant
ship built for Hiero II of Syracuse and was desigDed by Archi-
medes. It could fire a bolt 18ft (5.5m) long, or a stone weighing
1731b (78kgd up to a range of 200 yards (185m). To score a
direct hit on an approaching enemy ^p would, of course, take
some excellent marksmanship and a bit of luck, but if one did, the
stone ban would certainly smash right through the deck and hull
of any ordinary ship, and sink it very rapidly.

We must now go back in time, and trace another series of more
sophisticated developments which began about halfway through
the fourth century B.C. and went on at the same time as the de-
velopment of the big-bow stone-shooter — namely, catapults with
torsion springs. Direct evidence on these comes from the same

*Altertiimer von Pergamon (Berlin 1885-1937) X, 48-54, Plate 31.

Copyrighted material

CATAPULTS

107

andent writers, supported by a few illustratioiis, and has been
supplemented to some extent from another source — experiments
carried out on reconstrucdons. These have taken two contrasting
forms. Dr Marsden made, or had made for him, a number of

models of various types of catapult, some of them full-scale. He
followed the ancient sources as closely as possible, and his recon-
structions resembled the catapults in ancient illustrations very
closely. He was not, however, able to make authentic sinew-rope
(p. 109 below), and the performance of his arrow-shooter, though
impressive, fell short of that of the ancient weapons. At the other
extreme, a joint project between the Classics and Engineering
Departments at Reading University was run during the session
1971-2, and a catapult was designed and built. Owing to budget
limitations, authentic materials could not be used, and others,
including steel and nylon, were substituted, the design beii^ such
that the strength of the structure and the properties of the springs
matched the ancient weapons as closely as possible. As a result, the
final project had an appearance totally unlike any Greek or Roman
catapult, but a performance which ought to be comparable, and
some useful information has emerged from the study.

The basic design of the torsion spring was very simple. Tt con-
sisted of a bundle of strands of elastic material, with an arm or
lever thrust through the middle. When this arm was pulled aroimd
in a plane perpendicular to the strands, it stretched each of them
by an amount proportional to its distance from the soas around
vibidi the arm rotated. If die arm was then released, the strands
contracted and the arm flew back to its origoial position. Cata-
pults with one spring mounted hcmzontally and one arm were in
fact a much later development. The standard torsion-spring cata-
pult had two vertical springs, the arms swinging outwards hori-
zontally. Between the tips of the arms was a bowstring, and the rest
of the structure was essentially the same as that of a bow-catapult.
(Fig. 35.)

Clearly, the most crucial design problem was the choice of
material for the strands of the spring, which must meet a number
of requirements. Firstiy, it must not stretch too easily (in the
technical language, the Young's modulus must not be too low), or
the movement of the arms, limited to a 90° arc at the maximum,
will not generate enou^ tension. Secondly, it must be a material
that can be woven into a rope, smce it is very difiicult to anchor

Copyrighted material

108 ENGINEERING IN THE ANCIENT WORLD

the ends by any other means. Thirdly, it must be a material which
will remain consistently elastic, and not deteriorate under normal
weather conditions. And fourthly, (*Which', as the Rev. Mr.
Collins said in a different conteact, 'perhaps I ought to have men-
tioned earlier') it must not be too difficult to obtain, or too
expensive. Needless to say, an ideal material which fully met all
these requirements was never found. The nearest approach to it
was called, iii Greek, neuron and the next best was hair.

Fig. 35. The toision spring catapult.

Neuron is usually translated 'animal sinew', but its meaning is
not quite so precise as one might wish. It could certainly mean
muscle-fibre, or tendon, or a muscle with its tendons taken as a
whole. Only in later medical writers (e.g. Galen, second century
A.D.) does it come to have its modern meaning — a bundle of nerve-
fibres. Hero tells us {Bel, p. 1 10, Wescher) that the best material
comes from the shoulders or badcs of all animals except pig?, and
it has been discovered, he says (indicating some methodical testin^^
that the neura of an animal which get the most exerdse are the
most springy, such as those from the feet of deer or the necks of
oxen. The fact that he says *feet' rather than 'legs* suggests that he
is talking about the Achilles tendon, and not about muscle-fibre,
though a teudou is, of course, the termination of a bundle of
muscle-fibres, aud, as anyone knows who has prepared turkey legs
for the oven, the one sliades olT rather imperceptibly into the other.
The regular use of oxen as draught animals with a yoke fitted on
their necks, would account for the strengthening of the tendons in

Copyrighted material

CATAPULTS 109

that legion. A final, clinching argument in favour of tendon is the
simple fact that muscle-fibre is edible, and would have gone into
the stewpot. Only the tough, inedible, and otherwise useless tendon
would have gone into a war weapon.

How was it made into rope? Here, once again, we arc faced
with an almost complete lack of knowledge of a craft industry.
Rope-makers were very much in demand in antiquity, particu-
larly for making ships' tackle. They probably worked for much
of the time in the open air, and the average man in the street
would have paused to watch them on occasion. As a result, Hero
does not feel obliged to say anything about the process, apart from
mentioning a 'machine' (or perhaps a tool) for doing the job,
without giving any details. It was presumably the same as that fen*
making hempen or papyrus rope, with one di£Ference. The tendon
was probabty shredded or t«used into very thhi fibres, each of
which would have been shorter than the lengths of other materials
such as papyrus. The serviceable length of the Achilles tendon in
a large British cow is not much more than Sin (20cm), and almost
all the other tendons are shorter. How these strands were woven
into a rope which would not creep under very considerable tension
is still a mystery.

The Reading University project included a study of the proper-
ties of tendon (obtained from a local slaughterhouse) and a stress/
extension diagram is shown in Fig. 36. It will be seen that the
graph at low strain is not linear, but that beycmd a certain point
the curve straightens out at a much steeper angle. Other
researchers have shown that the pdnt at which this occurs, and the
eventual slope, depend on the age of the animal from which the
tendon is taken. The accepted explanation for the phenomenon
is that the minute fibres from which the material is built up have a
'crimped' structure, and that the behaviour at low strain repre-
sents the straightciiing-out of the crimp, while that on the linear
portion is elastic extension of the fibre itself. In their accounts of
how to fit a sinew-rope on a catapult, the ancient writers emphasize
strongly the need to pre-stress the strands. Though totally ignorant
of molecular structiues, they must have found by experiment that
they could greatly increase the power of the catapult by keeping
the sinew under some degree of tension even when the arms were
forward and the bowstring straight. Moreover, assessment of the
mechanical properties of tendon has shown that they did in fact

110 ENGINEERING IN THE ANCIENT WORLD

choose what wa8 probably the best material available to them,
smce its energy-storiiig capadty is, believe it or not, higher per
miit weight than spring sted. But there is one important qualifica-
tion which must be borne in mind. All these tests, and the numerous
other medical research investigations, have been concerned with
individual fibres or very small groups of fibres. When woven into
a sinew-rope their behaviour might have been quite different, for
two main reasons. Firstly, they are not then extended in a straight
line from end to end, but helically, and secondly, their adhesion to
each other, which is relatively unimportant in animal tissues (since
they are anchored at each end) becomes all-important in the rope.

Stress

Extension

Fig. 36

The other material used for torsion springs was hair. In most
contexts where it is mentioned, particularly those in which large
quantities are specified, we may assume that horse-hair is meant,
but the ancient vyriters do stress the merits of human hair, especi-
ally women's. There were several anecdotes current in the ancient
world about the whole female population of cities under si^^e
sacrificing their hair to make catapult springs. One of them con-

Copyrighted material

CATAPULTS

111

cemed the women of Rome in 390 B.a, and was certainly apo-
cryphal, as the Romans had no artillery at that date. This, how-
ever, did not prevent later Roman antiquarians from using the
story to explain the peculiarly-named cult of die 'Bald Venus'
(Venus Calva). Another, told of the women of Carthage in 148
B.C., may well be true.

On the treatment and maintenance of spring materials, as on
the making of the rope, we are told vktually nothing in the ancient
sources. Hero and Philo both mention that hair and sinew were
soaked in olive oil, and treated with grease or fat, both before the
rope was made and after it was fitted to the catapult. But Philo
points out (Chapter 61) that when sinew-rope is under tension
the oil which has been absorbed gets squeezed out, and if more is
put on, it will not penetrate. The ropes must dierefore be slackened
off before oil treatment can succeed. In another writer, a quantity
of pine-resin is mentioned in the same context as a consignment
of hair for catapult springs, but whether the two items have any
direct connection is not certain. If so, the resin, like the oil used on
sinew or hair may have served as a water-repellent — Philo notes
the destructive effects of rust forming on the iron parts in direct
contact \with the sinew, and it may also have served as a bonding
agent in the making of hair rope. General Schramm, a German
army officer who did considcraible research on ancient catapults
shortly before World War I, successfully used torsion springs made
of horsehair, but apparently nobody has yet made a usable sinew-
rope.

The second major problem was the design of the frames around
which the rope was to be wound.* To begin with, these were

apparently plain rectangular wooden frames with tenons on the
ends of the vertical members (which took the stress end-on) and
thick bars at the top and bottom. The spring-rope \vas anchored
at one end, then wound aioimd the frame, the other end being
tucked under some or all the layers at the top or bottom. At
this stage, the rope cannot have been put under much tension
except by keeping it taut tliroughout the winding process, and it
is di£^ult to see how this could have been done. To remedy this,
short iron rods {axonia) were put under the cords at the top and
bottom of the frame and, somehow or other, 'twisted' so as to

*This is a very brief summary of the story; for a fuller account, see
Maisden passim, ami particulariy HD, pp. 16-47.

Copyrighted material

112 ENGINEERING IN THE ANCIENT WORLD

increase the tension. Hero is a bit vague about this, possibly be-
cause it was akeady regarded as an obsolete method by the authors
he consulted. The next stage brought a big improvement.

A round hole was bored in each of the horizontal members of
the frame, and the spring-rope was threaded through, passing over
the iron rod which lay across the hole (Fig. 37) and then back
through again. This was repeated until no more strands could be
forced through the holes, even using a sort of monster needle in
the latter stages. Then the end was anchored as before (tucked

Fig. 37

under the other layers) and the springs were tensioned by rotating
the iron rods in the appropriate direction — clockwise for the top
left, anti-dockwise for the top right, and so on. This was a very
effective method, as it gave a considerable mechanical advantage.
But it soon brought out a serious fault in the design. As Hero states
{BeL p. 98, Wescher), the most powerful stress in the whole
machine acts direcdy on the iron rod, and pulls it against the
wooden frame. Even with very little tension, the friction thus
generated would make it extremely diflicult to rotate the rods, and
even if they could be turned, they would gouge out the wood,
especially at the points where they were parallel to the grain.

The simple answer to this problem was to put ^vashers between
the rods and the frame to take the wear. On the smaller machines
they were round, and made of solid bronze, with a top rim thick
enough to support the rods without denting; even so, the area of
contact between rod and washer would be less, and the friction

Copyriyhica niaiciial

CATAPULTS

113

between metal and metal much less. For the larger machines
bronze washers would have been heavy and expensive, so they
were made of wood, with the thrust end-on to the grain, and plates
of iron nailed onto the upper and lower surfaces. By this time, if
not carKcr, a sort of primitive spanner was being used to turn the
rods. Hero {Bel. p. 101, Wcschcr) calls it 'an iron lever (literally,
crowbar) with a ring, which engaged with the projecting part of
the rod', and it must presumably have looked something like the
implement shown in Fig. 38a. Using such a tool, there were ob-
viously two dangers. One was that the washer might shift sideways,
causing the outer strands of the spring-rope to rub on the sides of
the hole in the frame, producing fraying or abrasion. To prevent
it, two smaU holes could be drilled in the top of the frame dose to
the hole, and the washer could be made with two projecting lugs
which fitted into the holes, and stopped the washer from shifting
(Fig. 38b).

The other, more serious danger was that the rods, pressing with
great force on the top rims of the metal washers, might produce a
sharp burr on their inside edges, which might fray or even shear
the sinew rope. The answer to this problem is presented by Hero
as an alternative to the fixed washer with lugs; in fact, it was
probably a later design which superseded it. The washer was made
in the same shape, but with grooves in its top rim in which the
rods rested diametrically across the hole, and the washer itself
turned round with the rod. This, by the way, would make it easier
to use the spanner. To prevent Uie washer from sliding about
sideways, a 'groove' {soUn) was cut in the top of the frame. This
might have been a ample rebate, with the washer shaped to fit
into it, or (perhaps) a circular groove, concentric with the spring-
hole and of slightly larger diameter, with a projecting rim on the
under side of the washer which turned round in it (Fig. 38c & d).
This solved the latest problem, but brought back the earlier one
of friction between metal and wood, which made the washers
difficult to turn, and caused wear on the frame, with eventual
weakening. The answer to this, in turn, was to use a round metal
*under-plate' (in Greek, hypothetna) fixed on the frame, with the
Svasher-guide' cut into it. At some stage this seems to have solved
the friction problem so well as to raise the c^iposite one — the
washers tended to slip back and loosen off the tension when the
adjustment had been made and the spanner was removed. The

114 ENGINEERING IN THE ANCIENT WORLD

Fig. 38

designers succeeded in solving this problem and another, which
has not yet been mentioned, at the same time.

It was clearly necessary to make the thrust of the two springs
of a torsion catapult exactly equal. If, for instance, the right-hand
one pulled more strongly, the bowstring, and with it the tail of the
miasile, would be dragged to the right during projection. The
groove along which the missile travelled would check this to some
extent, but even so, the missile would almost certainly deviate
from ^e line of aim, to the left of the target. The simplest and
most obvious way of correcting imbalance (which could be de-

Copyrighted material

CATAPULTS

115

tccted by a number of simple tests) would be to tighten up the
weaker spring, or slacken off the stronger, by turning the washers,
and the engineers seem to have decided that it ought to be possible
to make this adjustment accurate to 7^°, or ^ revohition of the
washer.

This was the problem, and their solution of it was one of the
most ingenious ideas in catapult design. Our evidence for it comes,
not from Hero this time, but from archaeology. The remains of an
arsenal at Ampurias in Spain (in the extreme N.E., near the
modern Figueras) were discovered just before World War I, and
the finds, to be dated about mid-second century B.a, included
most of the metal parts of a catapult. Among them were washers
and 'underplates' of a new design. The wallers were shaped as
shown in Fig. 38e, with a flange which rested on the raised part
of the \md^late'. The slip-back was prevented, very amply, by
two pins, diametrically opposite each other, inserted through hoks
in the flange and in the underplate and — for a short distance at
lezist — into the wood below it. Now in order to get an adjust-
ment to the required accuracy, it might seem necessary to drill two
holes in the washer flange and 48 holes aroimd the rim of the
*underplate' at intervals of 7J°, leaving only very thin walls be-
tween the holes and, inevitably, weakening the 'underplate' very
seriously. But this was avoided by what might be described as a
very primitive Vernier system. 16 holes were drilled in the 'under-
plate' at a spacing of 22^°^ and three pairs of diametrically oppo-
site holes in the washer at a narrower spacing — 15^. Thus at every
step of 7^° — i.e. at 48 points per complete revolution — one or
other of the three pairs of holes in the washer would line up with
one of the eight pairs of holes in the 'underplate', and the washer
could thus be pinned in any one of 48 positions. Bearing in mind
that, to judge from illustrations, the washers were quite often given
as much as three complete turns during the tensionin.s: process, the
addition or subtraction of 49 turn would have been quite a fine
adjustment — roughly ±0.5%.

Simultaneously with the development of the washers and ten-
aoning apparatus came various improvements in the design of the
frames. To begin with, as already mentioned, the two spring-
frames were made separately, and attached to the sides of the
^ug^'. But this was a bad arrangement, as drawing back the
bowstring puts a very strong torque on each frame, tending to rip

uopyiiyhiea maiciial

116 ENGINEERING IN THE ANCIENT WORLD

it off its fixing at the front. Various remedies were tried, including
planks to join the two frames, but they increased the weight of the
machine without improving its power, and made it more difficult
to dismantle and re-assemble — a requirement stressed by the
ancient writers. After a short period of thought, the dc^gn was
simplified by the use of a single length of wood to form the tops of
both frames, and another for the bottoms. The stretching of the
springs then tended to push the whole frame, as a single unit,
backwards along the 'trough*, but it could be fixed by wooden
dowelsy or perhaps removable pegs.

Energy storage w created byextm
forward swin^ of arm

Breaking pdttt

Fig. 39

Tlie next concern of the designers was to increase the arc over
which the arms could move. To get the m2iximum energy storage
and release, it was necessary to make the difference between
maximum strain (with the arms drawn back) and minimum (with
the bowstring straight) as wide as possible, and one way of doing
this was to allow the arms to swing further forwards.**^

*The eneigy b represented in a graph of stress/extension (Fig. 39)
by the area bdow the curve^ between the vertical lines representing man-

Copyrighted material

CATAPULTS

117

Glosdy related to this was a problem which had already been
encountered. Though the tension in the springs had fallen off, the
arms were still moving quite rapidly at the end of their swing, and
carried considerable kinetic energy which had somehow to be
absorbed. It was soon found tliat if the arms struck the outer up-
rights of the frame they would be severely damaged. On the
other hand, if the bowstring was made shorter, so that it pulled
the arms up before they reached the frame, it would have to absorb
their energy, and would be suddenly and very violently stretched.
Machines adjusted in that way must have needed frequent re-
placement, both of the bowstring and of the arms which, when the
bowstring snapped, would smash themselves on the frame up-
lifts. Philo remarks (Chapter 68} that breakage of an arm
was one of the commonest accidents to befall catapults: it not only
put the weapon out of acdon for a long time, but was also highly
dangerous to the crew, as the broken portion flailed across in front
of them. This happened during a deliberate overload test on the
Reading project catapult, and amply justified the rigorous safety
precautions which had been enforced.

The answer to the problem came when the designers realized
that the torsion springs could, in another mode of operation, also
be made to serve as shock-absorbers, and take up some of the
energy which would otherwise have gone into the bowstring. This
was done by extending the 'butt ends' of the arms (which they
called 'heels', ptemat), rdnfordng them, and positioning pads on
the inner frame uprights so that the 'beds' struck them just before
the bowstring straightened. From then cm, the 'heel' would
act as a fulcrum, and some of the energy from the moving arm
would be used up in displacing the middle of the torsion spring
sideways.

As a result of this modification, the heel-pad became the limiting
factor on the forward swing of the arm. In order to increase that
swing, the inner upright was shifted backwards (i.e. away from the
'heel'), and the outer upright forwards. In order to give room for
this, while at the same time economizing on material, the top and

mum and minimum extension. The maximum can be pushed up (i.e. to
the right) provided it falls short of the breaking point by a reasonable
safety margin. If the minimum is then lowered (i.e. to the left) the area
(which represents stored energy) is increased, though this has progressively
less effect as the stress falls to lower values.

Copyrighted

118 ENGINEERING IN THE ANCIENT WORLD

bottom members of the frame were made with a convex rear edge.
Since the hed-pads required a certain amount of *give*, the earlier
problem recurred of the arms striking on the outer uprig^ts^ and
so a recess was cut out from the centre of thdr rear edges, just wide
and deep enough to clear the arm as it swung forward. To com-
pensate (partially) for the loss of strength which this caused, the
front edge of the upright was made convex (Fig. 40).

This des^ came to be regarded as the optimum for a modern
ately light, high-velocity arrow-shooter. It was known as 'straight-
spring' (in Greek, euthytonos) from its general appearance (it
should be remembered that the off-setting of the uprights is not
visible from above). They found, however, that in order to throw
heavier missiles it was necessary not only to increase the size of the
springs, but also to extend the forward swins: of the arms. This
was done by carrying the same method of off-setting still further,
until e\'entua]ly the front edge of the inner upright was made level
with the rear edge of the outer. The geometry of the design, as
given by Hero (BeL p. 94, Wescher) is as follows ; (Fig. 41).

Hcds

Fig. 40

Radius of arcs
VM£,ALC

sprlng'heU

Copyriyhica niaiciial

CATAPULTS

119

Though the difference between this design and the 'straight-
spring' 18 one of degree rather than kind, the term used for it re-
flected precisely that difference and no other: it was called a 'fold-
back spring' (palintonos). In practice, however, the much more
obvious difference lay in the fact that the *straight-spring' was
normally an arrow-shooter, and the 'fold-back spring', almost by
definition, a heavier and more powerful stone-thrower.

There is just one peculiar feature. Hero seems to imply that for
the palintonos they reverted to the old system of making the two
spring-frames separate, and used an array of planks and struts to
fix them to each other and to the 'trough'. He gives no reason for
this, but a simple explanation can be offered. The portion of the
diagram marked ALGEMD has to withstand the greatest stress
in the whole machine. If the grain of the wood ran parallel to DC
its load-bearing capabilities would be very limited, because almost
all the stress would fall on the cross-bonding between the fibres,
hardly any of which would be efTectively supported at both ends.
It would not take much tension in the spring to split the wood
along its grain. Tf, however, the grain ran parallel to AG and DE,
a fair proportion of its cross-section would be supported at both
ends. (The grain of the left-hand frame would have to slant in the
opposite direction.) As it was patendy impossible to find lengths
of wood in which the grain had two bends of about 26^ at the
right places, they used three separate pieces — hence the separate
frames. Why did Hero not explain this? Probably it was left
to the carpenter to p<nnt out what was necessary, and why
(*if you puts 'un on the slant to the grain, *t^^ill jigger 'un up,
m'dear'). Such vital steps in the development are not recorded in
our sources.

A very brief mention will have to sufhce for the other details of
these two types of catapult. Each was fitted on a tripod base, the
attachment (at or near the centre of gravity) being by a swivel
mounting (in Greek, carchesion, see p. 95) which allowed the whole
weapon to he tilted up or down and to swing sideways into any
position. The bigger stone-throwers had a beam-and-strut struc-
ture to support the Hrough', instead of a solid batten, which would
have been too heavy. Another feature very much emphasized by
all the ancient writers is that those pordons of the machine, parti-
cularly around the frame, which were subject to shock or wear
had iron plates nailed over them.

Copyrighted material

120 ENGINEERING IN THE ANCIENT WORLD

Something has already been said about the dimensions of bow-
catapults^ but on the toraon-spring types we have oonmderably
more information, thanks mainly to Phib and Vitruvius. These
catapults were classified in terms of the misdle they were designed
to shoot — the weight of the stone shot or the length of the bolt.
The latter may sound rather crude, but it is perfecdy clear that
arrow patterns were strictly standardized, and that for any given
length there was an accepted diameter and weight. This applied
equally to the very large bolts. Thus we hear of a *three-span
euthytonos^ — which was a straight-spring model firing a bolt
about 2 7 in (0.68m) long, or a 'twenty-mina palintonos* — a *f old-
back' spring model throwing a stone shot of about 191b (8.6 k§^.

The crucial dimension, as the ancient designers well knew, is
the size, and hence the power, of the springs. A theoretician might
measure this in terms of the weight dt anew-rope, but in the
practical directions given by Hero, Fhilo and \^truvius each spring
is regarded as a cylinder of set proportions, and the crucial
measurement is its diameter, which is taken to be the same as that
of the holes in the tops and bottoms of the frames through which
the spring-cord was threaded. All the authorities agree on two
formulae, one for each type of catapult. For a straight-spring
machine, the diameter D is pth of the length of the bolt, and for
the palintonos (stone-thrower) D = 1.1 WM X 100, where D is
measured in dactyls and M is the w^^ht of the shot in Attic minas*
Thus, for a three-span arrow-shooter the diameter D would be
= 4 dactyUy just over Sin (7.72cm), and for a twenty-mina
stone-thrower, just over lO^in (26.75cm).

Clearly, the first of these formulae does not involve any arith-
metical difficulty, but the second, without four-figure tables or a
slide-rule, is not so easy. It n simple to work out if M happens to
be 10 or 80 minas, since 1000 and 8000 happen to be cube
numbers, but what could the ancient artificers do when = 15,
or Af = 45? Three possibilities are mentioned by the ancient
authorities. Some workshops were lucky enough to have lists of
measurements for the various sizes of shot, worked out by a tame

*The Greek dactyl was very nearly | inch (0-76 in) or 19'3 mm. A 'span'
(spithame) was 12 dactyls and a 'cubit' {pechys) 24 dactyls — roughly 9 in
and 18 in (23 and 46 cm) respectively. The Attic mina (there were other
standards in me) was just tinder 1 lb (0*96 lb) or 436 g. One 'talent'
{takuOon) was 60 rninax, roughly 58 lb or 26 kg.

Copyrighted material

CATAPULTS

121

mathematician. The figures given by Philo and Vitravius were
derived from such lists. Failing thisy there is the cruder device of
finding (by trial and error) the nearest whole number to the cube
root: for example, if M = 45, 100 X Af = 4500, and the cube of
16 is 4096 and that of 17 is 4913. Having found this, they could
either split the difference or take 16 or 17 as though it were the
exact cube root, from motives which might be economic, or
personal, or even dishonest.

The third method, described by Hero and Philo, is much more
accurate. (Hero, Bel. p. 112-119 Wescher, Philo, Chapter 51-2).
It starts from die known spring-diameter of a successful catapult
For this purpose, let us take a ten-mina machine, where ilf ' X 100
= 1000, and Z) = 11 dactyls. It is possible to find the correct
diameter for a second machine (larger or smaller) by calculating
two mean proportionals (in Greek, mesai ana logon) between its
missile weight M" and that of the first machine, Af' (10 minas).
Suppose A/" = 40 minas, we then require two numbers, x and
such that \0 : X = X : y = y : 40. The simplest way of finding
these is by a geometrical construction, which would be within the
scope of all but the very dimmest artificer. This version is Hero's*
and Philo's differs only very slightly in detail. We draw two lines
AB and BG at right-angles, their lengths proportionate to M' and
M" (Le. in the ratio 10:40, Fig. 42). Cbmplete the rectangle
ABGD, and draw its diagonals AG, BD, intersecting at £.

&

C

Fig. 42

Copyrighted material

122 ENGINEERING IN THE ANCIENT WORLD

Produce D C and DA for some distance. Then place a straight-
edge, pivoting on B and cutting the extenaons of DA and DC
Rock the straight-edge about B until each of these points of inter-
section is the same distance from £, measured wi^ a ruler, and
call these points F and G (i.e. F£ = £6). It will then be found
that AF and CG arc the mean proportionals between AB and
BC— that is, AB : AF = AF : CG = CG : BC. It also follows
that the ratio AF/AB is the cube root of the ratio BC/AB, and
that the spring diameter suitable for Af' (1 1 dactyls) multiplied by
this ratio (AF/AB) will give the correct spring-diameter for •

TlUdbuff (n cm)

Length of anns

D = Duim eier oj' spring-hdts

TJilcknc^sct onUr
upriphi %D(75cni)
inner j^D(45an)

Thiekuess «D (Ucni)

Length qfttvugh^

Width at pm and ^
1wtt0mt&(Uem)

miff; af ctnUlt
'2D (z-^cin}

I'u'iifhtofuprighb

3'/D(42crn)

Total hciiiht of
frame (' 'fQi-ctive
working kngth cfsprtng)

Fig. 43

Once this basic dimension D has been calculated, the rest be-
comes easy. Both Fhilo and Vitnivius give a long list of other
dimensions, all expressed in terms of this diameter. Some minor

differences between their figures can easily be accounted for by
the developments one would expect to take place in the gap of
nearly two centuries which separates them. Here (Fig. 43) are a
few of the typical figures for a euthytonos (D = i of the length
of the arrow). Let us take D = 12cm. A corresponding set of
dimensions for the stone-thrower (palintonos) shows a few predic-
table differences. The arms are shorter (6D), but had a wider arc
of movement, so the 'trough' is longer (19/>). Tlie tops and bottoms
of the arrow-shooter's two frames are each made from a single
piece of wood, and its length is 6} D. The corresponding measure^
ment for the paUntonos is 2f D, vftkh is consistent with the sup-

Copyrighted material

CATAPULTS

123

pofiitioD (p. 119) that two separate frames were used on this
machine. Allowing 1^ i) for the width of the Hrough' between
them, the total width would be 6^ D, a litde more than the

euthytonos.

In conclusion, a brief account must be given of six develop-
ments in catapult design, four of them Greek and two Roman.
The four Greek ones — the repeater, the wedge, the bronze-spring
and the pneumatic — are all described by PMo. The first two may
possibly have been developed as working weapons, but the bronze>
spring and the pneumatic were almost certainly not.

The employment of the arrow-shooting catapult in ancient
times corresponded quite closely with that of the machine-gun in
modem warfare, up to the time (towards the end of World War 1}
when a really effective answer to the static machine-gun post was
developed, namely, the tank. The main function of the catapult
was to pin down enemy infantry. Troops approaching a fortified
position had either to risk hea\7 casualties or else move in small
groups, keeping to the 'dead giound' as much as possible, and if
the defenders' catapults had been skilfully positioned, this might
be very difficult. Conversely, the attackers could use their own
catapults to force the defenders to keep down behind the cover of
walls or battlements, where they could not take such effective
steps to prevent the scaling of the walls or the use of si^e-engines.
In each of these roles, the speed of fire was clearly an important
factor in the effectiveness of a catapult, and it is not surprising that
some attempt was made to convert the simple machine into what
nowadays might loosely be called an automatic weapon. In Greek
it was called a polybolos — 'multi-shooter'. It was developed in
Rhodes by one Dionysius, who came originally from Alexandria.
From this we may infer that there ^vas some degree of 'brain-
drain' going on between the various armament centres in the
Mediterranean area, and it is amusing to see how Philo, who
travelled around several of them, was on some occasions regarded
with suspicion, and peiiiaps deliberately misled by security-
conscious technicians.

The essential stepi in firing a catapult which, it would appear,
were normally carried out by a crew of at least two^ were as
follows : ( 1 ) push the slider (diostrdj forwards (2) lock the 'daw* on
the bowstring (3) draw back the diostra (4) place the missile in the
groove (5) take aim (6) pull the trigger. Dionysius succeeded in

uopyiiyhiea maiciial

124 ENGINEERING IN THE ANCIENT WORLD

automating (2) (4) and (6), and the business of ainimg was efTec-
tively eliminated by pre-aligning the catapult on a tripod and
firing it on fixed Ihies. The not4oo-difficult task of sliding the
didstra back and forth was speeded up by a chain-drive, so that
one man, with practically no training or experience, could use the
weapon.

The chain-drive deserves special mention, as it has possible
applications in other fields (see p. 73). This is how Philo describes
it (Chapter 75-6). 'It (i.e. the polyholos) does not have a cord for
pulling back the didstra, and instead, the capstan is made with five
facets on each of its projecting ends. Small blocks of oak (literally,
"briquettes", pUnthidia) with iron plates on each side are mortised
into each other and linked with pins, and these are wrapped around
the capstan. There are (chains of) these on either side of the
''trough", and each chain is fixed to the diosira (at one point) by a
pm with a round head/ The arrangement was as shown in Fig,
44a. Fhilo mentions handspikes (in the plural) on the capstan;
once again, it is surprising that a crank was not used.

The trigger, instead of a simple lever (see Fig. 44b) was in the
form of a rocker-arm, activated by vertical studs at one side of the
*trough', so that its working end was pushed under the back end
of the 'cla^v' when the didstra \vas fully forward, and was pulled
away to release the 'claw' just after the arrow had fallen into place
and the didstra reached its backward limit. The ammunition feed
was an ingenious mechanism, consisting of a long wooden cylinder
(something like a rolling-pin) which revolved inside a wooden tube
with slots at the top and bottom just wide enough for a bolt to fall
through. Over the slot at the top was fixed the magazine, big
enough to hold a considerable number of bohs, with sides sloping
inwards towards the slot (Fig. 44c). The cylinder itself had a groove
nmning from end to end, also just deep and wide enough to take
a bolt. When the cylinder was turned so that the groove lined up
with the top slot, a bolt would fall through into the groove. It was
then rotated half a turn, so that the groove came in line with the
lower slot and the bolt fell through onto the didstra, just in front
of the bow-string. The lower slot had just enough clearance above
the didstra for the trigger mechanism to pass under it. Philo is a
little confused about the bolts (or so it would seem). He says, quite
rightly, that they could not be notched, since they might not
happen to drop in the right position, but he also says that they had

CATAPULTS

125

Cai) The diain drive

Dicitm (dfrnrnJUUfback)

Loadintj
CYlindcr

Dlostra —

(b) Automatic

trigger

Vwstrnjbnvard,
lever Uiixist under
narofdaw^

Stud (fixed

Chtd:

Vwstra bach,
claw releases
bowstrinff

Stud-

Fig. 44. The rqieater catapult (pofybolos, 'multi-shooter*).

three flights p.e. they were arrows, not bolts — the same Greek
word is used for both). Even if this means that they had tliree
flights one behind the other and all in line, it is diflicult to see how
they could pass through the loading-cylinder or lie flat in the
groove on the didstra, if they did not happen to fall exactly right.
The rotation of the cylinder was also automatic. One side of the
tube was partly open, and the cylinder had a spiral groove on that
side, which ran halfway around the circumference, A short bar
with a knob on its end was fixed on the didstra to engage in this
groove, and this turned the cylinder so that the bolt-groove faced
upwards when the diostra was fully forward, and downwards
when it was almost f uDy back.

Having described this machine, Philo goes on to make some
disparaging conunents on it. His motives for doing so are not dear.

Copyrighted material

126 ENGINEERING IN THE ANCIENT WORLD

unless perhaps he could not remember some of the details accu-
rately (occasional remarks suggest this, and at one point he actually
admits it) and he might have been embarrassed if called upon to
construct one. His main critidsm is that it fires along fixed lines,
and is therefore useless against a moving target. Even firing at a
group might not be effective, because the beaten zone is short
and narrow. But this criticism would not apply when the catapult
was laid (for instance) on a crucial, narrow approach to a fortified
position. There are, moreover, two considerations which would
seem to override the objection. Firstly, like the Bren gun mounted
on a tripod, the catapult could be set up and aimed at an enemy
camp during daylight, and used in the darkness ivith devastating
effect. Secondly, esqperienced men who had been under fire from
ordinary catapults would be able to gauge with some accuracy the
time it took to wind back the slider and get ready to fire the next
bolt, but when encountering this type for the first time, they would
greatly over-estimate the 'safe interval' after each shot, and might
fall easy victims when they thought they were safe. However, Philo

does not seem to have appreciated this. His great worry, as an
artillery expert, is that unskilled infantrymen might get hold of
the weapon, blaze away indiscriminately, and waste ammunition.

The second of the Greek developments was an alternative
method of tensioning the springs on a torsion catapult. Philo him-
self may have been involved in the research, since he certainly
spends a great deal of time criticising the washer method (p. 112)
and expounding what he considers to be a great improvement.
(Chapter 56-67). What it all boils down to is that he returned
to the original, simple frame without spring-holes. Along the tops
and bottxmis of the frames were semi-cylindrical bars, to distribute
the pressure and wear over as much of the sinew-rope as possible.
Then the sinew-rope was wound around the outside of the frames,
and not pre-tensioned — or only a very little. Next, wedges were
driven in between the top members of the frames and the semi-
cylindi'ical bars, in effect expanding the frames and stretching the
springs. The two obvious advantages are (a) that for a given size
of frame more sinew-rope can be wound on, and (b) that the
bundles of sinew-rope are further away from the axis around
which the arms rotate, and so the leverage is greater. Philo claims
a number of other advantages, but he is not altogether objective
in discussing what might have been his own brain-child.

Copyrighted material

CATAPULTS

127

The third devdopineiit was the bronze-spring catapult, said to
be the invention of Gtesibiiis. Philo describes its design rather
cursorily, and adds a number of misguided comments on the under-
lying theory. Though he compares the properties of the bronze

used, in a rather confusing way, with those of 'so-called C'eltic and
Spanish swords' which were made of steel, there is no doubt that
the name 'bronze-spring' (in Greek, chalcotoiios) was an exact
term, and did not merely mean 'metal'. The alloy used had a
slighdy higher tin content than usual — 'three drachmas of tin',
he says, 'were added to each ndna of best quality purified red

bronze'. A drachma is (rather conveniently) one-hundredth part
of a mtno, so this would increase the tin content by 3%. Fhilo's
readers could be expected to know exactly what the normal tin

content of 'red bronze' w?is, but we unfortunately do not. Great
care was taken to obtain the best quality copper and tin, and to
purify them by repeated smelting, and the alloy was cast in the
form of two rectangular strips. Philo gave the exact measurements,
but the figures have not been preserved in our copies of the manu-
scripts. They were hammered t o the right thickness, bent around
a wooden roller, and then cold-forged for a long period, using
light hammers and light blows. According to Philo, the effect of
this was to case-harden them. If heavy hammers were used, the
springs would be hardened right through, and would become too
bitde. Then their ends were shaped and filed, and riveted together
to form an elliptical leaf spring (Fig. 45). One of these was
mounted on each of the outer uprights of an ordinary catapult
frame (a makeshift arrangement, which suggests an early stage

Copy I iyl uuU I Mdj-I lal

128

ENGINEERING IN THE ANCIENT WORLD

of research), and held in iron brackets, which also held the axles
on which the arms turned. A small projection on the 'hed' of each
arm (Fhilo calls it a 'bronze finger') thrust against the spring and
compressed it as the bowstring was drawn back (Fig. 46). Philo
then describes (with disapproval based on a completely faulty
theory) how Gtesibius attempted to increase the power by mount-
ing two such springs side by side on each arm.*

Philo claims, quite rightly, that bronze springs are less suscept-
ible to damage or deterioration through bad weather conditions,
but he also claims that they were more powerful than sinew-cord,
which is most improbable. The whole design has the appearance
of a tentative research project, and there is no evidence to show
that catapults of this type were ever in general use*

111'' .

Fig. 46. Bronze-spcing catapult.

Hie fourth Hellenistic Greek development was the pneumatic
catapult (in Greek, aerotonos, *air-spring'), also attributed to
Gtesibius. The description of this comes right at the end of Philo's
work, and is included, he says, *so that nobody may think that my
researches are incomplete' — an interesting and typical intrudon
of the public-relations angle. Whereas his account of the making
of bronze springs is in the first person (*We hammered the plates
. . . etc.'), that of pneumatic catapult is at second hand (*We were
told how Gtesibius demonstrated. . . .'). Glearly, we are dealing
with a tentative research project, which had apparently been
abandoned by Philo's time — about 50 years later.

The compressibility and elasticity of air had been known and
understood from quite early times, and a theoretical basis of a sort
was established by Strato of Lampsacus (third century b.g.). But
all ancient authorities seem to agree that Gtesibius was the first to

*Manden'8 interpretatton of this passage {T,T, p. 144) is incorrect,
and adds further conftnion to Fhilo's already confused thought.

Copyrighted material

CATAPULTS

129

make practical use of it in a number of mechanical devices, of
which the pneumatic catapult was the most impressive. Another
was the water-organ {hydrauUsX and Philo counters possible soepd-
dsm in his readers by pomtmg out that the piston-and-cylinder
pump on that apparatus, which they could see for themselves,
worked quite effectively.

The layout of the catapult was the same as for the bronze-spring
machine, except that pistons and cyhnders (with no outlets) were
substituted for the bronze plates. As the arms were drawn back,
some sort of tubular projections on their *heels' forced the pistons
into the cylinders, and compressed the air in them. When the
bowstring was released, the pistons popped outwards and swung
the arms forwards. Hie pistons and cylinders were made of
bronze, initially cast, and then hammered on the outside. The
cylinder, roughly shaped in the castmg, was bored out with some
predston, and the piston turned to fit it. Unfortunately, Philo
assumes that his readers know what tools were used and how,
and gives no details.

He does tell us, however, of a demonstration given by Ctesibius,
and reported by his informants, of the pneumatic 'spring'. The
cylinder was put into a clamp or vice, and the piston was driven
into it with a hammer and wedge. After a time, the air pressure
built up so high that even a firm blow from the hammer would
not force the piston in any further. When the wedge was knocked
out (sideways ?) the piston flew out with great force (and, we may
assume, a loud bang). The onlookers even reported that a flame
shot out with the air, a phenomenon which they e3q>lained as the
effect of air rudiing over the metal surfaces at high speed and
bemg heated by the friction. This explanation is a very familiar
one, being used by the philosophers as early as the fifth century
B.C. to account for lightning, and even if Philo's informants
imagined the whole thing, or made it up in order to impress him,
it is a fact that the cylinder would have heated up considerably as
a result of the air being compressed inside it.

It seems almost certain that this catapult never got beyond die
experimental stage. The reasons are obvious; to bore the cylinder
and turn the piston (a piston without rings, and with no gland on
its driving-rod) with sufficient accuracy was beyond the technology
0[ the time. If the piston fits too tightly, friction will prevent it
from moving fast enough and if it is not tight enough, the air

130 ENGINEERING IN THE ANCIENT WORLD

pressure will leak away. The best compromise between these faults
that could be reached in Alexandria in the third century b.o. was
not good enough, though surviving pumps £rom not much later
show that their skill was by no means negligible.

finally, two Roman achievements — the cheirobaUistra (liand-
catapult') and the onager (*wild ass'). Our evidence for the first of
these comes from the incomplete and rather problematic remains
of a treatise of that name by Hero, and from some illustrations,
notably those from Trajan's column. It was, as its name suggests,
a compact, portable arrow-shooter, with frames that were made
entirely of iron. Each torsion spring had two bars (corresponding
to the outer and inner uprights of the wooden frame) with rings
at the top and bottom in which the washers and tensioning rods
were fitted. They were held in position by two cross-members. The
lower one (to which the *trough' was fixed) was called the *littk
ladder*, being of a beam-and-stnit construcdon, and the upper
one the 'litde arch*, because it had an inverted U-bend in the
middle. The fixing of these components must have been strong and

rigid, though how it was done is not clear. Dr. Marsden suggested
welding, but that must have been quite difficult in a Roman forge,
without modern equipment. Riveting is an alternative, and was
used on Marsden's full-scale model (see T.T. pp. 232-3 and
Plates 6-7-8). The 'trough' and slider were of wood, and the
weapon was otherwise exactly like the older versions.

Its advantages over the older models were considerable. Though
probably not much lighter in weight, it was more compact, and
easier to set up — some illustrations show a semi-mobile version,
fitted in a sm^ cart. The sinew-brings were protected by metal
cases from enemy weapons and from the weather. In addition,
they were set further apart, which witii the shape of the Htde
arch*, gave a wider field of view and made aiming easier. The
fixing of the iron frames was so managed that the arms cotild
traverse a wider arc than those of a wooden-frame machine, thus
providing extra power and range. In this instance we have good
evidence (particularly from Trajan's Column) that the weapon
was fully developed and successfully used.

All the catapults so far described had two vertical springs, and
two arms which swung horizontally. What, then, of the onager^
that machine familiar from many cartoon drawings and classroom
models, with one horizontal spring and one arm which swings

Copyrighted material

CATAPULTS

131

upwards? There is remarkably little evidence for it in claasical
writers. None of the main treatises mentions it (Hero, Philo,
Vitruvius), and the earBest account which gives any detail at all

(and, by comparison with Hero and Philo, miserably little detail)
comes in the writings of Ammianus Marcellinus, who lived from
about A.D. 330-390. The surviving part of his history deals with
the years 353-378 a.d.

Fig. 47. The onager

The gist of his account says that the onager had two main hori-
zontal beams running from front to rear, made of oak or ilex,
which *rofle in a hump' near the middle, and were held apart by
cross-beams. A large hole was bored in each of the main beams,
through which ^e ropes' passed — we are left to guess that they
were sinew-ropes, tensioned with washers and bars in the same
way as the earlior machines. A wooden arm was thrust into the
middle of the bundle of ropes, with iron hooks on its tip to take the
sKng, made of tow or (rather surprisingly) of iron. As there was
no bowstring, and the arm apparently had no 'heel', some other
method of stopping it at the end of its swing and absorbing its
kinetic energy had to be clc\ ised. This was done by mounting a
large cushion, stuffed with chaff, on a braced structure above, and
just in front o£, the spring (Fig. 47). The arm was winched down
by two or more men, using a windlass at the rear end. Then the
shot was placed in the sling, and the sergeant in charge {magister)
'drove out the bolt which retams the whole rope-mechanism by a
mighty blow with a hanuner'. It is a pity that Ammianus could
not bring himself to cut down on the drama and give us more
technical details of the trigger mechanism.

From the silence of earlier authm we may infer that the onager
did not come into general use imtil much later than the two-spring

Copyrighted material

132 ENGINEERING IN THE ANCIENT WORLD

catapults, and from Aminianus that it reached its greatest popu-
larity in die third or fourth century a Why should this be so ? It
was probably a good deal easier and simpler to make than the
two-spring stone-thrower, and it did not require so much main-
tenance or adjustments such as the balancing of the springs, or
accurate alignment of the slider and trough. So long as the avail-
able technical skills were adequate for this, the two-spring machine
had the obvious advantage of better performance, but in later
days, when craftsmanship declined, the simpler and cruder
machine became preferable. In order to give a comparable range,
the single spring of the onager would have to be made very large.
On occasion it was apparently enormous — Ammianus speaks of
eight men being needed to wind down the arm, even with a wind-
lass. He also says that the recoil was tremendous (meaning the
reaction on rel^ise, which would thrust downwards on the rear
end of the machine, and might even cause the front end to jump
up) — so much so that it could not be positioned on top of a
masonry wall, because it would dislodge stones, 'not by its weight
but by the violent impact {concussione violent a)' . It was, he says,
used on a turf platform, which could absorb the shock.

The performance of the onager was probably improved quite
considerably by the use of the sling, and some interesting results
might come from more detailed research into the trajectory of the
missile with various lengths of sling and different adjustments of
of the hooks. This may well have been the method used to alter
the range — a longer sling gives a lower, flatter trajectory and
shorter range — but it is difficult to see how the direction of aim
could be altered, in view of the fact that the whole machine would
have to be shifted round. A f uU-flcale model* weighed over two
tons, and many of the ancient machines must have beien heavier
still. Only a very modest-dzed one could possibly have been
mounted on wheels or made efTecdvely mobile.

♦Described in the Appendix to The Crossbow by R. W. F. Fayne-
Gallwey (2nd ed., London 1958).

Copyrighted material

6

Ships and sea transport*

From very early times the lives of the Greek people have been
closely linked with the sea. Many of their city ates were chosen
mainly, if not entirely, because a good anchorage was available
nearby. Much of their import and export trade was carried by
sea, overland transport being slow and costly. It was navsd
supremacy in the Aegean that enabled Athens to dominate that
area for the greater part of the fifth century b.c, and to make
various communities, who had joined her as fellow members of a
defensive alliance, into the subjects of an Athenian empire. Tliat
same supremacy also enabled her to operate protected shipping
routes, and import large quantities of cereals ^om the Black Sea
area and elsewhere. This was especially important during the
Peloponnesian War, when Spartan invasions cut off access over-
land and destroyed the local crops before they could be harvested,
but importation remained vital during the following century and
later.

Preoccupation with die sea and ships often reveals itself in the

imagery used by Greek poets. The 'ship of state', with a political
leader as 'helmsman', became a cliche, and when Haemon (in
Sophocles' Antigone, lines 715-7) tells his father that 'a man who
keeps the sheets taut and refuses to slacken them in a squall will
finish his voyage keel uppermost', his words must have had a vivid-
ness and immediacy for the audience which they cannot have for
many of us today.

There were also, inevitably, important consequences for tech-
nology. For any city-state to maintain her dominance, it was
necessary for her ships to be superior not only in numbers but also
in design and performance. The princ^al technique of naval
fighting at that time was ramming, and its whole success depended

♦For the sake of simplicity all measurements in this chapter are
expressed in modern units: a table of Greek and Roman units and their
equivalents is given in Table 3 on page 169.

134 ENGINEERING IN THE ANCIENT WORLD

on speed and manoeuvrability; even a very slight advantage in
cither could make the difference between victory and defeat. So,

when city-states were willing to make financial resources available
and kept demanding new and better ships, it is not surprising that
a considerable body of expert knowledge was built up in the Greek
shipyards, or that ship design attained a high level of competence
and success.

The Roman approach was very different. At no stage in their
history were they really a seafaring people. Their capital city was
sixteen miles inland, with access to the sea (via the river Tiber)
only for small boats, and Ostia, at the river mouth, was quite
insignificant as a sea port until the major harbour works were
completed by the emperor Claudius in the mid-first century ajx,
by which time Rome had become dependent on the importation
of grain from overseas. As for naval strength, it was not until the
start of the Punic Wars in 264 b.c, when the Romans had to face
the Carthaginians — a nation with strong naval forces and a sea-
faring tradition — that they built their first fleet of warships. For
several centuries before that they had managed without one. It
says much for their courage and determination that they were
willing at that time to learn the task of building ships and training
crews from an army of landsmen.

Even so, at the end of the Second Punic War (201 B.C.) when
the Carthaginian naval strength was finally broken, the Romans
allowed their fleet to deteriorate without renewal, and when fleets
were needed in later times they were either raised from the Greek
ddes in southern Italy, or specially built. It was not until after the
Battle of Actium (31 b.c.) that the first permanent fleet was orga-
nized by Augustus, and permanent Roman naval supremacy in
the Mediterranean was established.

The Romans contributed little to the development of ship
design. They took over where the Hellenistic Greeks had left off,
with a preference for a type of warship which had been in existence
for more than a century. Their merchant ships also seem to have
been very much like those of their Greek predecessors. It is by mere
chance that a greater number of Roman illustrations have been
preserved.

An idea of the form of an early Greek ship can be gained from
two pieces of evidence — a passage from Homer's Odyssey (V,
228-61) and a cup by the vase-painter Exekias dated to the late

Copyrighted material

SHIPS AND SEA TRANSPORT 135

sixth century B.a* Fig. 48 is based on the cup illustration, with
some details added from other sources. It shows a pirate boat
which could be rowed or sailed; Odysseus' boat was for sailing
only.

The fact that each of these sources gives any kind of technical
information is in itself surprising. Odysseus has been detained on
an island by the beautiful nymph Calypso for a number of years.

Mast

Fig. 48. Early Greek ship

When he is at last allowed to leave she does not conjure up a boat
by magic or produce one which she has kept hidden. Instead, she
presents Odysseus with a set of tools, and shows him where to find

suitable timber on her island. Odysseus turns out to be not merely
the warrior hero of epic tradition, but also a skilled and knowledge-
able craftsman. Homer clearly assumed that his audience knew
about, and were interested in, the details of shipbuilding and
equipment, and Exekias, though he was illustrating a fantasy scene
from mythology, has taken care to draw the vessel correctly down
to quite minor detaiL

In this chapter, oared vessek will be discussed first, and sailing
vessels later. But first, an important difference between all types
of Greek and R(»nan boats and modem ones should be noted,

♦Arias-Hirmer-Shefton, A History of Greek Vase Painting (Thames &
Hudson, London 1962), plate XVI.

Copyrighted material

136 ENGINEERING IN THE ANCIENT WORLD

which is not obvious at first sight from the illustrations, but is made
quite clear by the Homeric accomit and from archaeological evi-
dence. It concerns the method of huU construction.

In recent tunes, most wooden hulls have been made by a method
called ^dinker* or 'lapstrake*. The keel is first laid down, with a
heavy, shaped beam (*kcclson*) on top of it, and the stempost and
sternpost jointed into it. Then vertical frames are attached at
intervals, which determine the eventual shape of the hull. The
outside shell of planks is then put on, starting with one either
side of the keelson ('garboards') each overlapping the one below
by a small amount. To keep these overlapping joints tighdy
together, it is necessary for the planks to run in one piece the full
length of the boat, and to be suitably springy.

This method of constnictum is characteristic of Northern
Europe, particularly Scandinavia, thou^ the time and place oi
its origin are obscure. The older method used by Greeks and
Romans, and in the Middle East until quite recent times, bdoi^ed
to the Eastern Mediterranean and may have originated in Egypt.
It is known as 'carvel' construction.

Instead of overlapping, the planks of the outer shell are jointed
edge-to-edge. In very early times they were held together by ropes
or threads passing back and forth through holes — as the Greeks
said, 'stitched together'. Ancient commentators on Homer, and
some modem scholars, have thought that he refers to this when
he makes Agamemnon warn the Greeks (in Iliad II, 135) that,
after some years on the beaches of Troy *the timbers of the ships
are rotting and the ropes losing their tension'. If so, this must be a
deliberate archaism, since Homer makes Odysseus build his boat
by a much more sophisticated method. Perhaps Agamemnon was
just referring to the ropes in general.

The later method of edge-jointing the planks, described in
Odyssey V, 246-51, involved a fair amount of joinery such as
could only be done by a skilled carpenter. At intervals of 3-9in
(7-22 cm) depending on the size of the ship mortises were cut into
each of the facing edges exactly opposite each other (Fig. 49)
and short wooden tenons made to fit into them and form a bond
across the join. These were cut with their grain end-to-end, at
right-angles to that of the planks. Their thickness was usually ^ to
i that of the planks, and their width varied according to the size
of the ship and the distance between them. When Homer says

SHIPS AND SEA TRANSPORT 137

□

Q

Fig. 49

that Odysseus 'did as much jdncry on his improvised boat as a
skilled craftsman would put into die hull of a broad merchant
ship', he simply means that he did the job thoroughly and without

skimping. Three methods were used to keep the tenons in place,
and thus hold the edges of the planks snugly together. First, they
were shaped to fit tightly into the mortises and were tapered slightly
towards each end, so they could therefore be called, both in Greek
and in Latin, 'wedges'. Secondly, some kind of waterproof adhe-
sive, made from resin, was used. Finally, to make quite sure that
the tenons would not slip out of the mortises, a hole was bored
through each plank and each end of the tenon, and round wooden
dowels driven through.

To hold the planks in the appropriate position for joining
together, and to give them the correct curvature, a series of 'timber-
cbmips' (in Greek dryochoi^ were fixed aroimd the keel during
construction, but they were outside the hull, and did not

Copyrighted material

138 ENGINEERING IN THE ANCIENT WORLD

eventually form part of the ship. In fact, the usual practice was to
shape and fit the boat-frames inside the hull after it was complete
or almost complete — the reverse order of construction from that
of a cfinker-buUt boat.

In order to strengthen and protect a hull made in this way when
the sea was very rough, the Greeks used devices called hypozomata
— 'under-belts'. The most famous mention of them comes in the
account of St. Paul's voyage (Acts, Chapter 27, v. 17). The gener-
ally accepted view (though there has been much argument about
it) is that the 'under-belts' were heavy cables, running along the
outside of the hull from stem to stem, which could be tightened
up during an emergency by means of windlasses. The modem
term for this jwactice is 'frapping*.

The shape and size of the hull varied greatly, according to the
function to be served by the boat, and a wide range of terms for
different types of vessel seems to have been used rather carelessly
and inconsistently. There were, however, two basic types of hull,
for which the simplest distinguishing terms were long ship' and
*round ship'. The *long ship' was essentially a warship or pirate
vessel, designed for rowing at high speed in action, though sails
could be carried for cruising or long voyages. The Vound ship'
was a merchant vessel for sailing only, apart from the smallest
ones which were rowed on ri\ ers or in harbours.

The differences in hull shape of the two types were exactly what
one would expect. The 'long ship' was slender, with a length-to-
beam ratio of about 10:1. A characteristic feature to be seen from
the side was the tall stempost, which curved upwards and for-
wards. Ancient ships were normally beached stem first, and in
many illustrations a short landing-ladder is shown tied to the stem-
post. At the bow another even more distinctive feature was a large
projecting ram, which looks almost like an eictension of the ked
and keelson. It could take various shapes. In older vase-paintings
it is like a rudimentary figurehead, but in the fifth century it
apparently had two chisel-like blades, one above and one below
the waterline. Their points were sheathed in bronze to increase
their destructive power. The join between the ram and the stem-
post was shaped to reduce water resistance, so that the whole
structure acted both as an armament and as a cutwater.

The *round ship' was much broader, with a length-to-beam
ratio of about 4:1. Since it was normally under sail, and in any

Copyriyhioa inaiciial

SHIPS AND SEA TRANSPORT

139

case unable to engage in fighting tactics> it had no ram. In order
to increase the cargo capacity the bows and quarters were full and

rounded, and there was usually a broad expanse of bottom which
was almost, if not completely fiat. The resistance on such a hull
as it went through the water was obviously greater, but that would
not be much of a problem at the sort of speed they could hope to
achieve.

Yet another distinguishing feature of ancient Greek and Roman
ships was the steering mechanism. Instead of a rudder, hinged cm
the stempost, they used a steering oar at the side of the hull near
the stem or (more often) one on each side. In illustrations through-
out the classical period the blades appear rectangular in shape,
and on *Iong ships' they usually trail in the water at an angle of
about 45°, projecting a little beyond the stern. If the ship had to
be beached stern first it was necessary either to remove the steering
oars altogether before grounding, or at least to hoist the blades up
into a horizontal position well above the waterlinc.

The method of mounting the steering oars has been the subject
of some argument, and it is difficult to reach firm conclusions from
the existing evidence. On warships, it appears that they had loops
of rope (^strops') attached at the appropriate pomt, which were
lowered onto some sort of vertical peg on the out^de of the hull
when the ship was launched, and shortly before landing, the oars
were lifted off their pivots and shipped. With this arrangement,
they could have been operated as in Fig. 50a, and one Greek
expression used for the helmsman's movements is 'pushing out'
and 'pulling in', which might refer to this. There is also, however,
evidence to suggest that on some vessels, particularly large mer-
chamtmen, the steering oars were nearly vertical and had their
round shafts mounted in a crude kind of bearing, in which they
were twisted round by means of tillers slotted into their top ends
(Fig. 50b). The tillers may or may not have been joined together
by a connecting rod, so as to turn together. This arrangement
Mwuld fit better with the statement in a passage of Ludan, to be
discussed later, that a diminutive and elderly helmsman was able
to control a very large merchantman and steer her into port with-
out any great effort.

One would naturally expect that the warship would require
more powerful steering apparatus than a merchantman, for two
reasons. Firstly, because it was longer and narrower, the thrust

Copyrighted material

140 ENGINEERING IN THE ANCIENT WORLD

Fig. 50. Mdiiodt of steering.

required to turn it around would be greater. Secondly, for the
purposes of ramming it would be necessary to turn in a very tight
circle on occasion, but if the crew were sufficiently competent, it
might have been possible for the rowers on one side to back water
and swing her round in that way. For a long time it was believed
that steering oars were less efficient and more difficult to control
than a modern rudder, but recent experiments with replicas of
Viking ships (which had a single steering oar, not unlike the Greek
and Roman ones) have shown that this view was quite mistaken.
Among other advantages, the larger steering oars on merchantmen
turned about a central bar, and were roughly symmetrical, with
the result that the water thrust evenly on each half, and in oppo-
site directions, which made the helm lighter.

The characteristic warship of the Greeks and Romans was a
'long ship' (as described above) propelled by oars during naval
actions. Mast, yard and a square sail could be used tor long sea

SHIPS AND SEA TRANSPORT

141

voyages, but for maximum peiformance they were normally
removed and left ashore when the ships were launched against
the enemy.

Apart from two modern refinements, oars and rowing techni-
ques have not changed much since ancient times, and a Greek
rower from one of the ships which fought at Salamis would find
himself much at home in a lifol)oat from the early part of this
century. One novelty he would notice would be the U-shaped row-
lock, which was unknown in antiquity. Greek and Roman oarsmen
had fixed vertical pegs (*tholepins') against which the oar was
pulled, and to which it was tied by means of a leather strap. The
Greek word for 'tholepin' was kleiSf meaning *ke/, and where
they are shown in illustrations, they are shaped as were early keys,
like an inverted L. Ancient writers on mechanics refer to the thole-
pin as the fulcrum in their lever analogies. If the oar VoUed' around
a circular peg, without rubbing too much, wear need not have
been too severe. There is also a remark in an ancient commentary
on a passage in Aristophanes {Acharnians 96) that the function of
the leather strap was 'to prevent rubbing on the woodwork', mean-
ing (presumably) the gunwale. This is consistent with the fact that
it was called tropos or tropoter, meaning 'tiu-ner' or 'twister'. In
addition to keeping the oar up off the gunwale, it kept it in position
under the horizontal part of the 'key', and also served to prevent
it from being lost overboard if the rower (through 'catching a crab'
or for other reasons) momentarily lost his grip on it. On the other
hand, it was not permanently fixed to the oar, being listed as a
separate item in the rower's kit.

Our ancient Greek would be used to rowing from a simple fixed
bench, and would be totally unacquainted with the sliding seats
of a modern racing boat : to prc\'ent blisters on the bottom, each
rower was issued with his own personal cushion.

Illustrations show an almost infinite variety in the numbers and
arrangement of rowers in various types of vessel, but in the case
of the oared warship it is possible to trace a continuous develop-
ment in design from 'Homeric' times (seventh century b.c.) until
the end of the classical period (late fourth century b.g.), culmina-
ting in the best-ever des^ of ancient warship — the trireme. After
that, development took a different line altogether, striving for size
and impressiveness rather than performance. Thm, in turn, came
a reaction against this trend, and a return to earlier, simpler

Copyrighted

142 ENGINEERING IN THE ANCIENT WORLD

designs, which persisted until the end of Roman sea-power in the
Western Mediterranean in the late fourth century a.d.

Homeric rowing ships had single banks of oars arranged sym-
metrically on either side. Small vessels for rajnd trandt, convey-
ance of despatches or important passengers had about twenty
rowers, and larger vessels, such as those which conveyed the Greek
contingents to Troy, had up to fifty. Such a vessel was called in
Greek a penteconter ('fifty-oarer'), and at a rough guess, it might
have been about 100ft (30.5m) in length overall and about one-
tenth of that in the beam.

From about the middle of the eighth century b.c. there appear
in vase-paintings ships with two superimposed banks of oars. This
change was accompanied by another, which must have been known
in Homer's time but is not mentioned by him — perhiq » because
he wished in this matter to maintain a 'genuine antique touch'. In
order to make it possible for ships to carry a landing-party, or to
fight at sea by drawing alongside and attacking each others* crews
(Homer has no occasion to mention either of these operations), a
raised deck was built, running the whole length of the ship but
not the full width. A space of 3 ft or so (Im) was left undecked
along either side to give headroom for the rowers and avoid the
danger of their being trapped below deck at the mercy of a
boarding-party.

The next developments are not well documented in literary
sources or illustrations, and the exact order in which they took
place cannot be established with certainty, but from the design
which eventually emerged we can infer what the key changes were.
When the second bank of rowers was added, it was probably above
the first, and since the original bank rowed at gunwale level, the
second bank must have been roughly level wi^ the raised deck.
Very soon, if not from the start, designers found that it was better
to arrange them 'in echelon' (Fig. 51a) than to have one directly
above another. Ihc fault of this design, however, is obvious. The
crew themselves made up a considerable fraction (about one-fifth)
of the total weight of the craft and, being positioned so high up,
they would cause it to be top-hea\7 and liable to capsize.

A better alternative arrangement was to place the second bank
below the first. To do this, a row of holes (*oarports') had to be
cut in the hull, and a reinforcing bar fixed below them to hold the
tholepins and take the thrust of the oars (Fig. 51b). This arrange-

SHIPS AND SEA TRANSPORT

143

mcnt lowered the centre of gravity and made the ship more stable,
but it caused another problem. If the sea was choppy, or if the
ship heded over during a sharp turn, water might be shipped

through the ports. To prevent this, a leather bag (in Greek,
askoma) was fixed around the oar and (somehow or other) around
the edge of the port, perhaps using the top half of a small aniinai's
skin, the oar shaft passing through the neck-hole.

Fig. 51. Arrangement of two banks of oan.

Eventually (the exact date is uncertain, but probably some time
in the latter half of the sixth century b.g.) these two designs were
combined to form a new type of vessel which became the standard
warship for more than a century. The Latin name for it was
tnretnis and this, anglicized as 'trireme*, is more commonly used
than the Greek form /nm J. Both of them mean (probably but not
quite certainly) the same tiring — a 'triple-rowed* or 'triple-oared'
v essel, that is, one with three banks of oars. There has been a lot of

Copyrighted material

144 ENGINEERING IN THE ANCIENT WORLD

controversy over the past eighty years or more about the design
and rowing arrangements, and it would be tedious and time-
CQDSuming to follow the history of the various theories which have
been advanced. In this context only the most recent — and best —
soludon to the problem, advanced by J. S. Morrison iu 1941 and
now generally accepted, will be discussed.

It is probable that in the prototype version the lines of rowers
in the three banks, viewed from the bow or stern, were vertically
above one another — the top bank level with the raised deck, the
middle bank at gunwale level, and the lowest bank rowing through
ports in the hull. This, however, would mean that the oarsmen of
the top bank would either have to use extra long oars, or else hold
them at a very steep angle to the water, and manipulate them very
much in among those of the lower banlEs. This problem was sdved
by building on an outrigger (called in Greek parexekesia —
'out-along oarage') which projected beyond the gunwale. The
rowers' seats were placed directly or nearly over the gunwale, and
the rail which carried their tholepins was about 2ft (60 cm)
outside the gunwale and perhaps a foot (30 cm) or even less, above
it.

There were a number of advantages in this arrangement. First,
the rowers in the top bank, because they were to one side ('out-
board') of those below them, did not have to be so far above them
vertically, and this lowered the centre of gravity making the ship
more stable without increasing its beam. Also, it enabled them to
use oars of the same length as those of the other banks, without
having to hold them at a very steep angle to the water. Even so,
their task was conadered the hardest, and they were on occasions
given higher pay than the others. They were called thranUaiy or
*stool-rowers*.

Those oil the middle bank were called zygioi^ or 'thwart-rowers',
and those on the lowest one thalamioi or 'hold-rowers', who had
the most unpleasant and dangerous position. If the ship got badly
holed, they were the most likely to be drowned or captured by
an enemy boarding-party. Also, as Aristophanes points out
with homely vulgarity {Frogs, 1074), they sat with their faces
rather close to the bottoms of the zygioi above and in front of
them. Their oarports were only about 18in (45cm) above the
waterline, and even with efficient oar covers, they must have got
quite wet.

SHIPS AND SEA TRANSPORT

145

No remains of ancient triremes have as yet been found by under-
water archaeologists, thou^ there may be a simple reason for this,
which is discussed later. But from illustrations, from inscriptional
evidence and from the remains of shipyards in the Peiraeus (the

port of Athens) it is possible to work out rough measurements for a
trireme, as siiown in Fig. 52, and also the number and arrange-
ment of the rowers. The last item of evidence has recently been

Fig. 52. The Greek trireme.

reinforced by the discovery of similar foundations in ancient
Carthage. There were 27 rowers on each side in the lower and
middle banks, and 31 in the top bank. Since the hull curved
gently upwards at the stern, restricting the space for the lower
banks at that end, it is probable that the extra, four thranitai of the
top bank were placed there, and not at the bow, one of them, close
to the helmsman, acting as 'stroke oar'. This made a total of 54 +
54 + 62 = 170 rowers, which, added to the helmsman, the boat-
swain (in Greek keleustes, the 'giver-of-ordeis'), the ship's com-
mander and junior officers, and a small party of marines, made
up a total crew of just over 200.

Looking back from the destroyers and motcn* torpedo-boats of
the twentieth century, it is easy to fall into the belief that ancient

Copyrighted material

146

ENGINEERING IN THE ANCIENT WORLD

warships were absurdly slow and clumsy. This was by no means
the case, but the reason why such a belief persists is that, until very
recently, no real attempt has been made to their perfor-
mance. To build replicas of a pentekonter and a trireme, and to
establish figures by experiment, would be prohibitively expensive,
but it is possible to make theoretical calculations of their propul-
sive power and water resistance, from which maximum speed can
be estimated. If, moreover, the same calculations come out right
when applied to a modern rowing boat of known and accurately
measurable performance, we may feel confident that they give a
close approximation to the speeds of ancient vessels. A brief outUne
of the method is given in the Appendix at the end of this chapter.

The estimate for the maximum speed of a pentekonter comes
out at 9.5 knots — just about the same as a modem racing eight.
Specially trained crack crews have reached 10.65 knots in still
water, but figures worked out from the Oxford and Cambridge
Boat Race are misleading, as that race is deliberately timed for a
favourable tide of 3-A knots. For the trireme, the figure is even

higher — about 11.5 knots. The sight of a fleet of these vessels
cutting through the water at such a speed must have been most
awe-inspiring, and the impact of their rams on enemy ships
devastating.

That, ho^vever, was the maximum speed which could be at-
tained by a well-trained crew in good physical condition, and
could only be kept up for perhaps ten minutes or so. If they had to
row over a long distance, they must have reduced their output
to about h.p. each, which would cut the speed down to just
below 9 knots. This was probably the sort of speed achieved in the
famous dash to save the Mitylenians, described by Thucydidcs
(III, 49). The Athenian CSoundl had passed a decree that all the
male citizens of that city-state (which had revolted against them)
should be put to death and the women and children sold into
slavery, and a trireme had been sent to convey the order to the
Athenian commander there. The next dav saw a chanoe of heart,
and a meeting of the Assembly revoked the decision by a small
majority. A second trireme was sent out immediately to try to
catch up with the first and countermand the order.

Thucydides does not tell us how long the voyage (of about 186
sea miles or 345 km) took, but he does say that the first trireme had
a start of *about a day and a night — say 24 hours — and that it

Copyrighted material

SHIPS AND SEA TRANSPORT

147

had been in harbour at Mitylene just long enough for the despatch
to be handed over when the second trireme arrived. The fint, he
says, *was not hurrying', and the crew of the second were promised
a large sum of money by the Mitylenean envoys and their friends

in Athens if they made it in time. They were provided with food
and wine, which they took while actually rowing, and (later on
perhaps) worked in relays, some sleeping while the rest rowed on.
There are two possible ways in which this might ha\ e been done.
If all superfluous gear was removed, and no personnel carried
except (say) one senior officer with the despatch and a few men
to prepare and serve the food, a spare crew for one bank of oars
could have been carried on board. 54 rowers would make <mly
about 20 'extra bodies' above the normal fitting complement,
and would not have slowed the ship down much. All three banks
could then have been manned all the time, the men rowing *f our
hours on and two off*, or some such arrangement. If so, they
could have kept up a speed of 8-9 knots and done the journey in
21-23 hours. By a lucky chance (says Thucydides) the sea was calm
and there was no headwind. The first trireme, cruising on one
bank of oars and not hurrying, might well have averaged as little
as 4 knots, and taken 24 hours longer.

Another possibility is that the normal complement of rowers
was carried, and each bank had one or more rest periods, getting
what sleqi they could in the bilges or on the deck while the other
two rowed on. This would reduce the speed to about 6.5 knots,
but if they all rowed for the first six hours and the last six, they
would still have made it in just over 24 hours. Navigation need
not have been a problem. If they set out early in the afternoon
(which is quite probable), they could ha\'e rounded Siinium and
reached the straits between Andros and Euboea by nightfall. From
then on they would be on open sea and could have steered by the
stars, keeping a sharp lookout for Psara if they went to the North,
or Psara and Chios if they sailed between. Sunrise would come
between there and the southern tip of Lesbos, and a six-hour stint
by all three banks from about 8 a.m. onwards would bring them
to Mitylene not long after midday. There were, after all, many
hundreds of lives at stake.*

*On the modern map by which these distances were measured (Hallwag,
Bern 1972) the trip by island ferry from Athens to Mitylene is marked
*about 20 hours*.

Copyrighted material

148 ENGINEERING IN THE ANCIENT WORLD

To describe in detail the tactics and methods of Greek and
Roman naval warfare is far beyond the scope of this volume, but
it may be helpful to list a number of limitations imposed on fight-
ing ships and their commanders by design and technical factors.

First, the trireme was not a *hcart of oak' vessel. It was made —
for lightness combined with strength — mostly of soft woods such
as pine and fir. These were not plentiful at all times in Greece,
and the Athenians imported large amounts mainly from Tlirace
and Macedonia. One result was that the hulls tended to soak up
water, and no effective surface sealing agent seems to have been
discovered, unless a substance called hypabipke' ('under-paint'),
listed among ships' chandlers' items, was some kind of varnish or
sealer. Consequently, all triremes were beached and man-handled
out of the water as often as possible. One of the most serious
problems for the Athenians in Syracuse harbour (as described by
Thucydides VII, 12-15) was that the enemy could launch their
ships at any time they chose, whereas the Athenians, having no
reserve of vessels, had to keep all of theirs in the water all the
time in case of sudden attack. As a result, their hulls had become
waterlogged, and they could not make anything like their maxi-
mum speed. The shipyards with sloping slipways at Zea in the
Peiraeus and at Carthage \\ ere partly for shipbuilding, but also for
drying out and cleaning the hulls.

Secondly, the trireme was designed for speed at the expense of
stability. Its performance in calm waters was very high, but in
anything like a roug^ sea it was neither fast nor safe. This meant
that it was rarely possible for warships to maintain a sea blockade
over a long period. The chances were that a stormy spell would set
in, during which the heavier, slower merchant ships could sail in
(provided the wind was in the right quarter), but the risk of
launching warships against them would be too high. There were
also other reasons, which are mentioned later.

Thirdly, being made as light as possible, they tended when
holed by the enemy to settle in the water without actually sinking.
The Greek word kaiadueirty which is almost invariably translated
as 'sink', in fact means no more than 'dip' or *lower' and when a
Greek writer wishes to indicate that something Vent to the
bottom', he generally uses a different word. So, when triremes
were holed in a sea battle, though they had become absolutely
useless as fighting vessels, the ccmibatants went to great lengths

Copy I iyl uuU I Mdj-I lal

SHIPS AND SEA TRANSPORT

149

and some risk to recover the wrecks. For example, in the engage-
ments off Naupactus in 429 b.g. (described in Thucydides II, 83-
92) some Athenian triremes were driven ashore and damaged —

quite seriously, it seems — by the Corinthians, who then tried to
tow them away against opposition from the Athenian allies on the
shore. Later, when the tables were turned and the Athenians got
the upper hand, they recaptured the damaged ships and towed
them back to their naval base. This mav also account for the fact
that no actual remains of ancient triremes have yet been found by
underwater archaeologists. Such woodwork as they have fomid
has come from merchant ships, and has been carried to the sea
bottom by the weight of the cargo. It could be argued that the
heavy bronze casing of the ram would be likely to drag a holed
ship under, but this is certainly not the impression one gets from
eye-witness accounts. If the ram broke off, it would of course sink
rapidly on its own.

From this and from other evidence we can appreciate that
another advantage of trireme construction was that it lent itself
to extensive repairs and re-fitting. When Polybius describes the
loss of a Roman fleet in a storm off Sicily, and adds that the
damage was so great that the wrecks were not worth recovering,
he obviously regards this as exceptional This in turn also accounts
for the fact that a number of ancient warships — both individual
vessels and fleets — seem to have been in service for a remarkably
long time. About twenty years was quite normal for a trireme,
but we do not know how many re-fits it would undergo in that
time, or how much of its original woodwork would have been
replaced.

Another vers- important consideration is that, however well
designed the trireme may have been, it was only one half of a
partnership, the other being a fit, well trained crew whose morale
was high. Ancient historians repeatedly stress that it was difficult
to make a ramming attack in narrow waters. The target vessel
had to be seen and selected from a distance, and the approach
course and speed had to be very finely regulated. The 'window*
during which an effective strike could be made was very short
indeed, as is shown in Fig. 53. Assuming the modest speed of 9
knots, each ship would travel its own length in about 6^ seconds.
If tiie attacker arrived about 4 seconds too soon 4), he could
himself be struck and holed by the target vessel, and if he tried to

150 ENGINEERING IN THE ANCIENT WORLD

AUadti^ vessel A

AttadUingvessU
eady (t-4)
ccrrcctcd. by

Steering

^ ^/ I Attacking
y / I vessel

I yearly

Target vessel
(cmrsmmUngthki. GSseconds}

Fig. 53. Ramming.

avoid this by turning to starboard, his ram would strike only a
glancing blow, and inflict very little daniagc. Up to pci ha])s four
seconds after the optimum time [t to t-\--\) some damage, though
less than the maximum, could be inflicted. After that, the target
vessel was virtually safe, since the speed of impact (being the
difference between the speeds of the two vessels) fell off rapidly,
and the attacker could deliver no more than a mild bump. In fact,
he would do much better to give up and turn to another targiet
It is clear, therefore, that during ramming manoeuvres the com-
mander and helmsman would have to rdy on the all-out effort and
total loyzdty of every member of the crew. A mere handful could,
if they so wished, upset the rhythm of a whole bank, and make
accurate steering and speed control impossible.

The clifTiculty of finding enough crewmen of the right calibre
was a perpetual problem for all ancient navies, and the shortage
(more or less permanent) of trained crews made itself felt in various
ways. One of the less attractive features of Greek naval warfare
was the not uncommon practice of slaughtering oarsmen captured
on board enemy vessels or picked up from the water. Later navies,
particularly the Roman, tended to rely on bigger, slower vessels,
and used boarding techniques in preference to ranmuqg, which
meant fewer rowers in proportion to the 'marines'. The bludgeon,
one might say, took over from the rapier.

Copyrighted material

SHIPS AND SEA TRANSPORT

151

Finally, the trireme was equipped simply and solely for rowing
and ramming. There were no officers' calnns, crew's quarters,
food store or galley. In fact, more than one naval engagement was

won by the commander of an attacking force choosing his time
skilfully so as to catch the enemy crews ashore, procuring or
eating their main meal of the day. And when large-scale expedi-
tions or troop reinforcements were sent overseas, those who
travelled in the warships (and not in the supporting fleet of
merchantmen) must have had a very crowded and uncomfortable
trip. Also, this made it impossible for triremes to stay at sea for
long periods — e.g. to maintain a blockade: they needed a base
nearby to which they could return at least once each day.

The history of the oared warship after the trireme is extremely
complicated, and beset by lack of evidence and much controversy.
It will therefore be summarized very briefly.

The next developments were the *quadiircmc' and 'quin-
quireme' (in Greek tetreres and penteres respectively). The trireme
had three banks of oars, and it might be thou<^ht that these vessels
had four or five, but there is no evidence for any ancient warship
having had more than three. The probable explanation is that a
*four-rower' had four men in each *box' or unit of oarsmen, just
as the trireme had three — one from each bank, the thranUai^ the
zygjioi and the thdUamoi as described above. If so, there must have
been more than one man on some of the oars. The most probable
arrangement for a quinquireme was two on each oar of the upper
and middle banks and one on the lowest. Apart from a short-lived
fashion for much bigger vessels, this became the standard warship
in succession to the trireme in Hellenistic Greek navies, in the
Carthaginian navies of the fourth and third centuries B.C., and in
the Roman navies down to the end of the Republic (31 B.C.). By
comparison with the trireme it was broader in the beam and had
a deeper and a bigger displacement — of the order of 75 tons as
against 40 for the trireme, which enabled it to work in rougher
weather, and to carry more fighting troops on deck — as many as
120 on the Roman ^ps used in the Second Punic War. It had
the additional advantage that a certain number of more or less
unskilled rowers could be taken on, by makmg each of them work
alongside a skilled and experienced companion. It was, however,
considerably slower and more cumbersome than a trireme.

Our evidence suggests that the quadrireme was introduced

Copyrighted material

152 ENGINEERING IN THE ANCIENT WORLD

early in the fourth century B.a and the quinquireme soon after.
About the middle of that century the 'sixer' was developed —
presumably with two rowers on each oar m each banL Then the
development accelerated rapidly. After the death of Alexander
the Great in 323 B.a an 'amis race' sprang up between the so-
called 'Successors' — his senior generals, who divided his kingdom
and resources among themselves. Almost every denomination —
the 'sevener', 'eighter' and so on — up to a 'thirteener' is in evi-
dence by the end of the fourth century, and by 288 B.C. we hear of
a 'fifteener' and a *sixteener*. The ultimate in these battle-ships
came with a 'twenty-er' and two 'thirty-ers* built by Ptolemy II in
the second quarter of the third century.

In the absence of any real evidence for the disposition of the
rowers in these big ships, scholars have speculated on various
possible arrangements — some of them credible, and others less
so. There is one practical lunitation, which applied as much then
as in medieval and Renaissance galleys. A long oar pulled (or
pulled and pushed) by a number of men — known as a *multiple-
rower sweep' — cannot be more than a certain length. This is
simply because the men near the inner end can only move over a
limited distance backwards and forwards or up and down, and
when they have to stand up, or climb up on a sort of ladder, to get
the blade of the oar into the water, their efficiency is much re-
duced. In Renaissance galleys eight men to a sweep was found to
be the biggest practical number.

Some of our sources also tell us that the bi^;est battleships were
*double-prowed' and 'double-stemed', and had two helmsmen.
The best (though not the only possible) interpretation of this is
that they were large catamarans with a single solid deck supported
on two hulls. The unsolved question is whether there were, or
were not, rowers on the inward-facing sides and if so, how much
space they had, and how they managed to row effectively.

The drive to produce more and more massive ships had by now
spent itself, and most later Greek fleets consisted mainly of quin-
quiremes, with just a handful of larger ships ('niners' or *seveners')
used by naval commanders as flagships. There was, however, one
final, futile effort made by Ptolemy IV of Egypt in the last quarter
of the third century B.C. — some 50 years after the heyday of the
big warship. This was a f orty-er' (in Greek, tessarakatUeres\ a
description of vdiich was written up by Gallixenus of Rhodes, who

Copyrighted

SHIPS AND SEA TRANSPORT

153

lived in the first half of the next century, and may have seen the
vessel still afloat His remarks are quoted in part by Plutarch
(Demetrius 43.5) and Athenaeus (V, 203e-204b). Even allowing a
lot for exaggeration (and this from natural scepticism, not from
any conflicting evidence) the specifications are truly astonishing:

Length overall

425ft

130.5m

Beam overall (or each

hull?)

58ft

17.6m

Height from tip of

sternpost to water-line

80ft

24.5 m

Length of stceiing oars

(4 in all)

45ft Gin

13.8m

Length of oars in top bank

(the longest used)

57ft 8in

17.5m

Athenaeus goes on to say that, *in the course of a test', the follow-
ing personnel were put on boai d ;

Oarsmen more than 4,000

Other ratmgiB 400
Marines (on the deck) 2,850

Total 7,250

Plutarch, and a number of modem scholars, have taken these
figures to represent the normal crew, but it seems much more
probable that they were derived from an experiment designed to
see just how many men could be put on board before the ship got
dangerously low in the water. It suggests that the total displace-
ment might have been of the order of 1 ,000 tons.

Given the personnel on deck to man a huge array of catapults
and missile-droppers, and to put big boarding-parties on to any
captured vessel, such a warship was without doubt unapproach-
able, unsinkable and altogether invincible. There was just one
slight problem — it was also practically immovable. According to
Plutarch, it was just a showpiece which stayed at its moorings and
never went into acdon, and although here (as so often) he is com-
plaining about the extravagance of the rich, he may be right.

The evidence on ancient merchant ships is mainly literary and
inscripdonal, supplemented in recent years by the discovery of a
number of wrecks. There are a few ancient illustradons, but this

Copyrighted

154 ENGINEERING IN THE ANCIENT WORLD

type of evidence is almost completely lacking over the last five
centuries B.a, which is in many ways the most interesting period.
We are therefore forced to rdy heavfly on the assumpticm that ship
design was conservative, and that Roman reliefs of the second
century a.d. give us a fairly accurate picture of the HeUenistic
Greek ships of 400 years earlier, and of those of the classical period
before them.

It has already been remarked that almost all ancient cargo
ships were under sail. Only the small craft were rowed, such as
lighters for bringing in cargo from a big vessel moored off -shore,
or for river traffic. There is very little evidence for combinations
of sail and oar power, except in warships which appear to he
trying desperately to get away from an engagement There are
good technological reasons for this : the requirements of design
for an oared vessel conflict on a number of points with those for
a sailing vessel, and vice versa. So, if both power sources are used,
one or other is being misapplied in the wrong type of ship. The
heavy ^aleasses of the Renaissance represent an attempt to com-
piuiiiisc bclwccii the conflicting requirements, and as such are
considered to ha\'c been a failure.

The most striking contrast between an early Greek and a modern
sailing boat is in the sail itself. Until late antiquity (probably the
fourth century A.D.) almost every vessel in the Greek and Roman
world, so far as we can tell from illustrations, had a square sail
set athwart the hull, imlike the 'fore-and-aft' rig, in line with the
keel, which is standard nowadays. A square sail in its crudest
form, like a modem spinnaker, is only effective when the wind is
blowing from dead astern or nearly so, but even the earliest
illustrations show three methods of controlling the area and
position of the sail, by which it could be made much more adapt-
able.

Odysseus' home-made boat, being smallish (he had to launch
and sail it single-handed) may have had a fixed mast, but the
larger vessel in which he set out on his adventures, Uke the other
Greek ships in the Trojan expedition, had a mast which could be
lowered onto a support at the stern ('crutch') when the ship was
beached, or in a calm when it was being rowed. On launching,
the mast was hoisted up into position and held there by two ropes
called protonoi ('front-stretchers') running from the mast-head to
points near the bow on either side. These ropes combined the

Copyriyhioa inaiciial

SHIPS AND SEA TRANSPORT

155

funcdons of a *forestay' and 'dirouds*, preventing the mast from
bending backwards or sideways. If diey snapped (this occurrence
18 described In Odyssey XII, 409-12) die mast could fall into the
Stem and strike the helmsman. It was supported against a strong

following wind by the 'braces' described below.

The foot of the mast rested in some kind of socket on the keel.
When it was hoisted into position ('stepped') and the stays were
made fast, the horizontal yard from which the sail hung was
hoisted by its central point, the sail having been furled up close to
it We gather from Homer that in early times the ropes used for
this purpose ('halyards') were made from plaited ox-hide thongiB.
It is unlikely that pulleys were used much before the fifth century
B.a — the surviving remains of them are from much later. Before
that, smooth metal rings might have served instead. The position
of the yard was controUed by two ropes Cbraces', in Oeek Hop-
ropes', hyper ai) running from its ends down into the stem. It
could be set square for a following wind, or swung around either
way to a position approaching that of a fore-and-aft rig, its
movement being limited by the protonoi. It could also be tilted
down at one end and up at the other. On large merchantmen in
later times it thus served as a crude substitute for a crane for load*
ing and unloading.

To spread or furl the sail, and to control its effective area, a set
of light ropes ('brails') was used. These were attached at one end
to the foot of the sail at intervals of about 1-1 ^ft (30-45 cm).
From there they ran vertically in parallel lines up the front of the
sail, being held in portion there by rings sewn on to the sail. Then
they passed up over the yard and down into the stern area. In a
number of pictures one man is shown holding about half of them
as a bunch in one hand, so clearly, it did not require much of a
pull to furl the sail. It may even have been necessary to hang
weights on the foot of the sail to make sure that it spread itself in a
very light breeze when the brails were slackened.

The third method of controlling the sail was by means of two
strong ropes ('sheets', in Greek 'feet', podes) running from the
bottom comers of the sail. On a small boat, if the wind was not
directiy astern, the one on the windward side was usually made
fast to a deat, while the hehnsman held the other (the 'lee sheet^
in his hand. Pulling this rope taut has the effect of catching the
full force of the wind in the sail, and to do so in a squall is to court

Copyrighted material

156 ENGINEERING IN THE ANCIENT WORLD

disaster — hence the remark in Sophocles' play quoted at the
start of this chapter.

The Mediterranean and its weather placed severe restrictions
on any seafaring, particularly under sail. Having no magnetic
compass, Greek and Roman sailors navigated by the stars at night,
and by landmarks in daylight, and so, in anything less than *vcry
good' visibility, they were very much at risk. Storms are more
frequent and severe in winter, but keeping to the summer season
does not remove that danger altogether. The earliest Greek lore,
preserved by Hesiod, recommended that voyages should be made
between June 21st and August 10th — a very curtailed season,
even allowing for the fact that he was a fanner, who had made
only one long voyage (which he had not enjoyed) and had all the
pessimism of his profession. More adventurous seafarers regarded
late March or early April as *a bit risky, but possible', from the
beginning of June to mid-September as We', from then until
early November as 'doubtful' and the rest of the year 'definitely
ouf , except for dire emergencies. The Garrulous Man, number 3
of Theophrastus* Characters, among various other platitudes
about rising prices and the weather, is wont to say 'Just fancy —
the sea was navigable from the Dionysia (late March) onwards!'
For most of five months in every year, the entire conmiercial life
of trading ports effectively closed down.

The later developments in the design of merchant ships show
two or more auxiliary sails — a small square one on a sort of bow-
sprit projecting beyond the bow (called artemon in Greek) and a
triangular one above the mainsail ('topsail*, in Latin siparum). But
at all times the main propelling force came from the single, big,
square mainsail. The performance of a vessel with this type of rig
in various wind conditions has been the subject of a good deal of
argument, and it would seem that some practical e3q>eriments,
which could be carried out without great difficulty or expense,
might enable us to get a more accurate picture.

It is generally agreed that Greek and Roman sailing vessels,
given a good following wind, could make speeds of about 4-5
knots, with 6 knots being exceptional, but not impossible in ideal
conditions. It should be remembered, however, that such evidence
as we have tells us the average speed over quite a long voyage —
of the order of 300 nautical miks or more — and may conceal a
wide variation of qpeed between i^urts and lulls. A squar&rigged

Copyriyhioa inaiciial

SHIPS AND SEA TRANSPORT

157

vessel could sail quite effectively, though not so fast, with the wind
on dther beam (at rig^t-angles to the desired course, Fig. 54a).
Tim was achieved by using the devices described earlier. The yard
was braced around aslant to the wind, and the 'sheet* on the wind-
ward side was let out so that the wind carried it forward of the
mast. The other ('lee') sheet was drawn tighter (a winch would
have been used for this on the bigger ships) and used to control
the sail in such a way that the main thrust on the mast was forward.

Fig. 54

But however well this was done, two ade-effects were inevitable.

The vessel would drift sideways through the water, off course to
some extent, and wonld also *head into the wind' — that is, turn
clockwise if the wind was on the starboard beam — and correcting
this by the steering oars would not be easy. In fact, the ancient
remedy is described in the Mechanical Problems attributed to
Aristotle. Apparently the brails were used to shorten up the lee-
ward half of the sail (the part astern of the mast) thus reducing
the area of sail directly onifronting the wind, and enabling the
forward half, which tends to turn the ship in the opposite direction,
to counteract the 'heading^ tendency.

The brails were also used when the wind was extra strong. If it
was due astern, the middle of the sail was farailed up, leaving only
a small area spread at each end of the yard. If a really sharp squall

158 ENGINEERING IN THE ANCIENT WORLD

blew up suddenly, the whole sail could be furled quite quickly,
leaving only the mast and yard to catch the wind, and this could
be done from the deck, without any of the crew having to *go aloft*,
as they had to in later sailing ships. Ahematively, it is sometimes
said that the yard itself was lowered part-way down the mast, but
this seems a clumsy and slow manoeuvre by comparison.

So far we have dealt with conditions in which the wind was
'following', or on the beam. When it blew from ahead of the beam,
things became more difficult. The action described above, of
swinging the yard around and adjusting the sheets, could be
carried a little further, so that the ship actually sailed slightly into
the wind, but how far this could be tsJcen is a matter of doubt and
guesswork. The usual approach to the question is to start from
the capabilities of square-rigged vessels of the eighteenth and nine-
teenth centuries. These are known to have sailed, with some diffi-
culty, what is variously called ^six pdnts off the wind' or *two
points into the wmd' — that is, at the angle shown in Fig. 54b. On
the assumption that Greek and Roman seamen were less com-
petent, it is usually said that they could only manage *one point
into the wind' — that is, with the wind just over 11° forward of a
line at right-angles to the keel (Fig. 54c). It would be difficult to
challenge this assertion without experimental evidence. What is
certain is that they would have tried their utmost to improve the
performance of their ships in this respect, for reasons which will
become clear.

What happened when the wind was too far ahead for the
braces and ^eets to cope is very obvious from a number of pass-
ages in Greek and Roman authors. The ships resorted, as saiUng
ships have done ever since, to 'tacking*, which involves setting a
course alternately to the left and right of the destination (the 'port
tack' and 'starboard tack' respectively). If the wind is dead ahead,
this means a symmetrical zig-zag course, and it is at once
obvious that here is a very slow way of travelling. Even in a
fair breeze, a ship sailing close to its maximum angle into wind
cannot make its full speed, and it has to cover more than five times
the direct linear distance. In theory, geometrically speaking, it
makes no difference how many tacks are made — four short tacks
from A to B cover the same distance as two long ones, but in
pracdce there are a number of other considerations. Hie business
of changing from one tack to another is hard work, and involves

Copyrighted material

SHIPS AND SEA TRANSPORT

139

a lot of organization, and the vessel slows down and loses time
during the manoeuvre; naturally, therefore, the tendency is
to make as few tacks as possible. But the room may be limited in a

channel or strait, or if the ship is following a coastline, and the
helmsman does not wish to lose sight of land. Again, it is difficult
to make changes of tack in daikness or poor visibility. It is also
unwise to go out to sea on a long tack if there is a likelihood of the
wind changing before the completion of the second tack. All these
considerations have to be weighed against each other, and in the
ancient world, without even a compass or rudimentary charts^ it
must have required great skill and long experience to make correct
decisions on the best course.

If the wind is not dead ahead, the tacks become asymmetrical
and in some conditions the second tack actually takes the ship
away from its destination, though this is more than offset by the
fact that the first tack is longer.

So far, it has been assumed that one point (11°15') into the
wind was the maximum capability of an ancient sciiiare-rigged
ship. It is clear, however, that even a very slight improvement on
that would make a marked difference to the distance travelled
and the time taken. For instance, with the wind dead ahead an
improvement of just one d^ee — too small to be detected without
quite sophisticated apparatus — would shorten the total tack dis-
tance by 8%, and a further improvement of 1^ would shorten it
by 15% altogether. If the ancient navigators could have managed
one-and-a-half points into the wind, the total distance would have
been shortened by ^, which would have made a big difference on
a long voyage against the wind, when the ship was tacking for
most of the time, and many such voyages were regularly made in
the ancient world.

For example, from Italy (Brindisi or the Straits of Messina) to
Alexandria was a 'downhill run'. The prevailing wind during the
sailing season was (and still is) N.W., and, if it blew steadily,
merchant ships could make the journey in 18-20 days, at an
average speed of just over 2 knots. According to PHny {Nat. Hist,
19, 3-4), a small, fast sailing boat could make it in 9 days, which
represents an average of about 4^5 knots. The return trip, how-
ever, was very different. They had to beat into the wind for ahnost
the whole voyage, making it much longer in distance and time —
anything between 40 and 65 days, or even more. The longer times

Copyri

160 ENGINEERING IN THE ANCIENT WORLD

probably indicate a number of weatherbound delays in harbour,
rather than a very slow rate of sailing, but even so^ a very slight
improvement in tacking performance would have shortened such

voyages by days, or even weeks.

An account of a voyage on precisely that run, made by an
unusually large merchantship, is given in Lucian's dialogue The
Ship, or Human Wishes^ written in the second century a.d. The
opening passage is clearly intended as a parody of the opening of
Plato's Republic, but the description of the ship and its voyage are
introduced entirely for their own intrinsic interest, and the satirical
genre does not call for any distortion or exaggeration in that part
of the work.

The ship was named the 'Isis'^ and was normally on the
grain run from Alexandria to Rome. Its measurements are given
as:

and from these its cargo capacity has been estimated at about
1200 tons. Even so, it was steered into port by 'a little old man,
who turned the huge steering oars with a slender wooden rod ; he
had curly hair, receding at the front, and his name was Heron'.
All this sounds (as Ludan intended) very circumstantial.

They had started from Alexandria in a light breeze — appar-
ently a little W of N W, and sailed roughly NNE, sighting Acamas
(Gape Amauti, the western tip of Cyprus) on the seventh day.
This is about 250 sea-miles, and represents a speed of about If
knots — quite reasonable for a heavy vessel on a port tack in a light
breeze. Then things went disastrously wrong. A westerly gale blew
up. Though they would probably have been making for Anem-
ourion (*Windy point', now Anamur on the S. coast of Turkey),
they were instead carried 'aslant' {plagioi, at right-angles to their
intended course) and ended up at Sidon (in the Lebanon, about
20 miles south of Beirut). The change of wind direction need not
have caused this change of course, but presumably the gale caught
them before they had rounded the tip of Cyprus, and was too
strong for them to beat against. The captain (wisely, no doubt)
ran eastwards before the gale, and perhaps went further south
than he needed to (about £S£), because he could not navigate in

length overall 182ft
beam *more than' 45 ft

height from deck to bilges 44ft

55.5m

13.9m+

13.4m

Copyrighted material

SHIPS AND S£A TRANSPORT

161

the bad viability. Then, still in rough weather, they went north-
wards around Gape St. Andrew and westwards between Cyprus
and Turkey (this, to Greek sailors, was Aulon, 'The GhanneF) and

got to *S wallow Islands' (now Gelidonya) 10 days after leaving
Sidon.

The coast in that region is very dangerous, with jagged rocks
and big breakers. One of the earliest wrecks so far discovered,
dating from the Bronze Age, was found there. In addition, they
ran one of the gravest risks for ancient seamen — they came on
this coast in pitch darkness. But, for a change, they had some good
luck. They sighted a fire (perhaps a lighthouse or warning beacon)
which told them they were nearing Ismd, and 'one of the Dioscuri
(Gastor and Pollux, the patron deities of mariners) set a bright star
on the carchesion and steered the ship to port when it had been
driven close to the diif . This is an obWous reference to 'St. Elmo's
Fire' — the static-electric brush discharge which appears on the
tips of wooden masts and spars in an electric storm. Carchesion
in the context of a ship normally means 'masthead', but the pheno-
menon, wherever it may have appeared, was taken as a divine
admonition to turn to port (out to sea) rather than stay on course
or turn to starboard.

Then, when they had been blown so far off the normal course,
the captain decided to give up the trip to Rome, and make for
Athens instead. Ifis reasons are not given — perhaps the caigo had
begun to deteriorate. Even then, the ship had to 'sail aslant into
the Etesian wind' — that is, tack up the Aegean heading against
the seasonal Norllierlies — and it arrived in the Peiraeus 70 days
after leaving Egypt. It should have 'gone to the south of Crete, to
the west of Malea (a dangerous area then as now) and should have
been in Italy (i.e. Ostia) by that time'.

This was, quite clearly, an exceptionally big merchant ship,
though not unique by any means. Hiero II, king of Syracu5;p from

about 265-215 B.a had a 'super-freighter', named the SyraciLsia,
built under the supervision of Archimedes — or so the tradition has
it. A detailed description of this huge vessel by Moschion, a near-
contemporary historian, has been preserved for us by Athenaeus
(V, 206d-209b). It carried a mixed cargo of the order of 16-1800
tons, and also a most formidable array of weaponry and a contin-
gent of over 200 marines, since fiGero presumably wished to make
it both proof against pirates and capable of blockade-running. In

Copyrighted

162 ENGINEERING IN THE ANCIENT WORLD

the event, however, it sailed to Alexandria, the only port in which
it could be safely berthed, and was there presented to King Ptolemy
III, the father of the one who built the biggest-ever warship. So
perhaps that rather absurd project was an attempt to 'keep up
with the Hieros'.

Ships of this size would clearly present technological problems
in building, repair and maintenance. So far as the construction
went, it seems that the components of a smaller ship were simply
scaled up. In the Syracnsia, for instance, the pins holding the hull
planking to the frames were made of bronze, some weighing 9Mb,
and the biggest 14^ lb (4.3 and 6.5 kg). These corresponded to
ordinary nails in a small boat. Similarly, the tenons for edge-
jointing the planks of the hull were longer and wider.

Whereas a small boat could be hauled ashore for maintenance,
the bigger merchantmen had to stay in the water once they were
launched. In fact, then as now, the hull was only part-built before
launching, and finished on the water. This raised serious problems
of maintenance. The hulls would not only become waterlogged

and leaky, but they would also suffer from that scourge of wooden
ships, the naval borer {teredo navalis as the Romans called it, and
zoologists still call it) — the marine equivalent of woodworm or
the death watch beetle. Ancient shipwrights avoided using certain
woods for the hull, because they were thought to be suscepuble to
it, larch particularly so, according to Pliny {Nat. Hist. 16, 79).
But for their bigger ships they adopted a drastic and expensive,
but effective remedy. The hull was covered on the outside, first
with a layer of linen cloth soaked in pitch, and then with what are
called in the ancient sources *lead tiles' — square or rectai^ular
pieces of thin sheet lead, pinned on the outside. This technique is
described by Athenaeus (V, 207b) and the evidence of no less than
eight wrecks, ranging from the fourth century B.C. to the first a.d.,
fully bears out his statement. The cloth layer would improve water-
proofing, while the lead layer would improve it still further and,
presumably, poison the borers and likewise barnacles or other
shellfish which tried to attach themselves. Underwater main-
tenance would thus be reduced to a minimum.

It is a curious fact that no apparent reference is made in ancient
soiu^ to 'careening' — the technique of tilting ships over to a
steep angle, so that the hull on one side of the keel is accessible for
cleaning and repairs. The Greeks and Romans certainly had avail-

SHIPS AND SEA TRANSPORT

163

able the cranes, miches and tackle required for this^ but if it was
common practice, they seem strangely reticent about it.

There is, however, clear evidence for a much more imprcsave

and sophisticated method, used on Ptolemy's monster warship.
Owing to the corruption of a single word in the text of Athenaeus
(V 20 1 c-d) the whole account has always been interpreted as re-
ferring to a slipway for construction and launching, despite the
fact that it makes complete nonsense as such. What we have here
is without doubt a dry dock, crude but workable, for repairing and
re-fitting the vessel. A rectangular trench, the same length as the
ship and slightly wider, was dug close to the harbour at Alexandria.
A foundation layer of squared stone was put down on the bottom
at a depth of just over 7ft Gin (3.5m). On top of that, wooden
cradles were laid crosswise from side to side, leaving a clearance of
just over 6 ft (1.85 m) above them. Then a channel was dug
through to the harbour, the trench filled up with water, and
the ship was towed in *by locally recruited casual labour'. Then the
channel was blocked up again, and the water pumped out of the
trench 'by means of organa\ the word regularly used for the force-
pump. Thus the ship 'settled safely and firmly onto the cradles'.
If it was in fact a catamaran, the whole operation would be much
easier and safer. As far as logisdcs are concerned, such a dock,
with the ship inade, might have contained something like | million
gallons of water (3.4 X 10*1) and the pumping-out operation
would have required some 500-600 man-hours.

Evidence for the cargo capacity of ordinary Greek and Roman
merchantmen comes from various sources, and requires careful
interpretation. For instance, a regulation forbidding Roman sena-
tors to engage in trade included a clause to the effect that anyone
operating not more than two ships of about 12-15 tons burden
each was 'not engaged in trade for the purposes of the Act'. This
has been taken to imply that vessels of about that size were regu-
larly used, but in fact the exemption level was probably put very
low indeed, to discourage attempts at evasion of the law. From
certain harbour regulations which have survived, it appears that
ships of less than 70-80 tons burden were not allowed to clutter
up the wharves in those ports or use the facilities, and when the
emperor Claudius tried to improve the Roman com supply by
offering favourable interest rates for shippers and what amounted
to free state insurance against loss by storm, the lower size limit

uopyiiyhiea maiciial

164 ENGINEERING IN THE ANCIENT WORLD

was placed at about 65-70 tons burden per vessel In other words,
vessels smaller than that were not considered to be making a signi-
ficant contribution to the transportation problem, and no wonder,
as we shall see shortly.

An ordinary, smallish merchantman seems to have had a cargo
capacity of 120-150 tons, and at a very rough guess, it might have
been about 60ft long (18.3m) at the vvaterhne and about 20-25 ft
(6-7.6 m) in the beam. Ships of 400-500 tons burden were by no
means uncommon, and we do not know how many very large
ones were in service. It is significant that those singled out for
special mention are all well over 1,000 tons.

Cargoes were stacked on ships in various ways, as evidenced by
underwater archaeology. Grain was usually in sacks, each contain-
ing something like a bushel, and liquids were mostly in amphorae^
the characteristic tall, necked jars from which wrecks of Greek
and Roman ships are often first identified. Heavy materials such
as stone or metal were stowed on a floor just above the bilges, and
they, and the amphorae, were packed in place with brushwood.
From some wrecks it appears that there were three or four floors,
each with its own layer of amphorae, one above the other. Some
grain ships had bins in the hold, into which the grain was shot in
bulk, which would save time on loading, but would add a good
deal to the unloading time, as the grain would probably have to
be put in sacks to be carried ashore.

Of the countiess journeys made by a handful of ships from port
to port, carrying anything from marble for building, metal ingots,
oil, wine or grain to looted art treasures, we have very litde written
record. Contracts were often drawn up between the merchant, the
ship owner and, in some cases, the banker who put up the money
for a trading voyage (and, through a clause cancelling the debt if
the ships were lost at sea, provided the ancient equivalent of ship-
ping insurance). But such documents w^ould be destroyed when the
transactions were complete. Our main evidence consists of refer-
ences in speeches (written for private lawsuits) to the usual terms
and conditions. One of these speeches, preserved because it was
believed to be by Demosthenes himself {Against Zenothemis, No.
32) tells of a curiously modem-sounding swindle which involved
faking an 'accident' to the ship, but which misfired.

But there was one unique and quite unparalleled transport
operation carried out during the early centuries of the Roman

Copyrighted material

SHIPS AND SEA TRANSPORT

165

Empire which was on such a huge scale (by the standards of the
time) that some account of it survives. Until about the middle of
the first century B.a Rome was able to supplement her home-
produced com supplies from the Western parts of the empire —
from Sicily, France, Spain and North Africa. But with the annexa-
tion of Egypt after the Battle of Actium (31 b.c.) a new and much
greater source was secured (very carefully, as it was under the
personal control of the Emperor) for the p^rowing urban popula-
tion. According to the Jewish historian Josephns {Bell, Jud. II,
386) it supplied one-third of Rome's annual needs. The grain was
grown in the Nile valley and the upper Delta (mostly in areas
inigated by the seasonal Nile floods), taken down river to Alex-
andria in hundreds of small boats, and then stored in grain-silos
or loaded on a fleet of grain ships. On a conservative estimate, the
total quantity exported in normal years was in the order of 130-
150,000 tons.

Owing to the restricted sailing season it was impossible for the
grain ships to make two complete roimd trips from Alexandria to
Ostia. If about half the fleet wintered in Ostia, they could sail (in
ballast, or with light cargoes) at the very start of the season — early
April — and get to Alexandria by early May, There they would
load up, and sail to Ostia in (say) 65-70 days, arriving before the
end of July. If unloaded promptly (this was not always managed)
they could leave again late in August, and reach Alexandria before
the end of the season. Those which had wintered there would sail
at the earliest pos^ble opportunity, and arrive fully laden in Osda
perhaps early in June. If the turn-around was quick enough, they
could leave by the end of the month, get to Alexandria by late July,
and re-load. But then it might be touch-and-go whether they got
back to Ostia before the bad weather set in.

Since the ships required docking and repair facilities during the

winter, it would seem reasonable to keep about half the fleet in
Ostia and half in Alexandria, but because the second trip to Ostia
was not always possible, there would be a tendency for more than
half of them to end the season in Alexandria. However, since that
meant a greater number available for the double trip the next
year, it would not present a problem. The total shipping require-
ment would be about two-thirds of the tonnage to be carried —
say 90,000-100,000 tons. How this tonnage was made up, we have
no idea. A possible fleet might consist of about twelve big ships of

Copyrighted material

166 ENGINEERING IN THE ANCIENT WORLD

1,000 tons burden, 160 of half that size, and 32 of 250 tons. That
would be the absolute minimum, leaving no maigin at all for
delay or loss at sea, or delays in cargo handling.

The first grain ships tacking up the Gampanian coast in the
spring must have made a brave sight, and Virgil, land-lubber
tiiough he may have been, gives a graphic description of Aeneas'
fleet sailing across the same stretch of water {A.6Ti6id V, 827—32) i

And now a sweet joy steals across Aenea^ mind;
*Step the mast/ he orders, 'spread the canvas alo ft !'
Attset their sheets alike, and braced the yards

Swinging and slanting now to port, now to starboard,
Brailing up the sail to leeward with each change oj tack.
Scurrying along in the fair breeze. . . .

Appendix to Chapter 6

METHODS OF ESTIMATING THE MAXIMUM
SPEEDS OF OARED VESSELS

The propulsive force of a pentekonter and trireme can be esti-
mated on the basis of various assumptions. One is that Greek
roivers had the same sort of bodily strength as an ordinary fit man
(not a trained athlete) today. This enables him to generate about
i h.p. for a short period (ten minutes or so) and about h.p. more
or less indefinitely. Some of this enegy is lost because oars, though
better than propellers, cannot be made more than about 75%
efficient even with the aid of modern science. If we assume that
ancient oars were about 70% efficient, this will probably be an
error on the low side. The maximum propulsive power for a pente-
konter would then have been about 11. 7 h.p. (about 6,500 ft lb/sec
or 8,700 watts) and for a trireme about 40h.p. (22,000ftlb/sec or
3 X 10^ watts).

From these figures it is possible to calculate the total thrust on
the vessel (applied at the tholepins) at various speeds. Since the
power is the product of thrust X vdodty, the thrust diminishes as
the speed increases, so the graph of thrust against velocity is a

Copyrighted material

SHIPS AND SEA TRANSPORT

167

hyperbola (Fig. 55). Also, as the speed increases the resistance
encountered by the ship increases, so it will gather speed until the

point is reached at which the falling thrust and the rising resist-
ance reach the same le\'cl, and exactly cancel one another out —
that is, where the two graphs cross in Fig. 55.

0 S 10 ts

speed knots

Fig. 55

The water resistance is made up of two components — skin
friction and wave-making reastance. The first is, in effect, caused
by the water Vublnng across' the surface of the hull where it is in
contact. To reduce it to a minimum it is very important to ensure
that the hull surface is smooth, and even a very slight roughness
(a grazed patch on the hull or the odd barnacle here and there) can

Copyrighted material

168 ENGINEERING IN THE ANCIENT WORLD

have dramatic effects. It is also important to keep the *wetted
surface' as small as passible. On a pentekanter it was about
620sqft (57.5m') and on a trireme about lOOOsqft (92.8m').
Wave-making resistance, which occurs when the boat throws

up a wash, can, like the skin friction, be regarded as a reverse
thrust on the hull which opposes that of the rowers, but the two
thrusts are related in difTerent ways to the speed of the vessel. Skin
friction resistance rises slowly in a steady curve (Fig. 55) roughly,
but not exactly, as the square of the velocity. Wave-making resis-
tance is almost negligible up to a certain speed, rises in a sharper
curv e above that, and finally shoots up to an enormous amount.
The point at which these changes of r^;ime occur are determined
by the speed of the vessel in relation to its length; this factor, the
*ielativfr speed* or *Froude*s number* is given the symbol ® and

V

can be worked out very simply from the formula y'=^> where V

is the sp>eed in knots and L the length in feet (at the waterline).
When © is below 0.7 wave-making resistance is very slight and
can be neglected. Above that it rises in a steep curve, and when ©
reaches about 1 .3 it becomes very large. For the pentekonter and
the trireme these values of © would represent:

^«0.7 ©=1 ©=1.3
Pentekonter L= 85ft 6.5 knots 9.2 knots 12 knots
Trireme L = 100ft 7 knots 10 knots 13 knots

Another factor has to be taken into account here — the fullness'
of the hull. This is calculated from a sort of Platonic Idea of a Full
Hull^ which is a cube based on the waterline length — in effect, the
beam, draught and length all equal. The fullness of a real hull is

its volume displacement divided into that ideal cube. For a long,
slender ship such as a racing eight this comes out at a very small
figure, but for a short, tubby barge it is much higher. Fullness has
not much effect on the critical speed at which wave-making resis-
tance begins to rise significantly, but it does affect the steepness of
the rise beyond that point, in practical terms this means that Greek
warships could cruise at speeds below 6 knots with almost no loss
of energy through wave-making, and in fact, the trireme could be
rowed at such speeds with only one bank of oars manned. Their
maximum speed would be somewhere between 6 and 12-13 knots,
and the fact that each comes out near the top end of the bracket

Copy I iyl uuU I Mdj-I lal

SHIPS AND SEA TRANSPORT

169

18 a tribute to the success of the design. Finally, it was totally
impractical for anyone in the ancient world to try to increase that
maximum beyond 12-13 knots by oar-power. A graphic confirma-
tion ci this can be provided by comparing a pentekonter and a
World War II motor torpedo-boat ('M.T.B.'). These two vessels
had about the same length and about the same displacement (20
tons or so). The M.T.B. had twin diesel engines delivering
l,500h.p. altogether, but this enormous reserve of power — more
than 90 times that available in a pentekonter ^ enabled it to achieve
only 4J times the speed (42 knots). In that case © is about five,
and the energy wa^ed in wave-making is very high indeed.

Table 3. Greek and Roman measures of capacity.

I2kyatlim

c fxcstai

1 kyatlios
1 xestes
1 chous

LIQUIDS

cyathus
sextfirius

0.06 pt

0.96 pt
5.76 pt

0.045 I

0.545 /
3.27 /

congnif

{1 Roman
ampliora or
quadrantal

r 1 Greek amphora
^ ai ap hoireiMi metretes

A Roman amphora jar weighed about 37^-39^ lb (17-18 kg) w hen empty. Filled
with wine (5| gall or 26.17 /) it would w eigh about 951b (43 kg). Thus 24 Roman
amphorae, or 16 Greek, would weigh about a ton.

8 congii s

12

5gaU6.08pt 26.17/

8 gall 5.12 pt 39.25 /

6 kyathoi

1 B

Greek

1 kyathos
kotyle
or hemma

1 xestei

SOLIDS
Roman
cyathus

0.08 pt
0.48 pt

0.045/
0.27/

0.96 pt 0.545 1

lgaU7.36pt 8.731
llgaU4.16pt 52.86/

2kotylai

16 aextarii « — 1 modhis

96 xertai 1 mfdiinnoii —

1 modiiit B jiiit under \ bushd

I medinuuw radier less than 1^ budids

Weights of wheat, according to Fliny (JVdtf. Hist, XVm, 66)— i.e. grains alone,

after threshing.

Light ■] r20 librae - 14.4 lb - 6.55 kg.

Medium >pcr modius^ 20f librae = 15.03 lb = 6.82 kg.
Heavy J \^21} librae = 15.7 lb = 7.12 kg.

On average grain, therefore, one ton = about 150 modii or 25 medimnoi.

Copyrighted material

7

Land transport

The shortness of this chapter by comparison with the preceding
one reflects — crudely but quite accurately — the unimportance of
land transport by comparison with sea transport in the classical
world. Any large community which was not self-supporting in
foodstuffs grown locally, or raw materials produced locally, de-
pended on importation, which invariably meant importation by
sea ; two outstanding examples — Athens during the Peloponnesian
War and Rome under the early empire — have already been
quoted. No such transport operation could possibly have been
undertaken by land; even the transit from Ostia to Rome of the
grain imported from Egypt, Libya, Spain and elsewhere, was
mainly by barge up the Tiber, in preference to road vehicles, des-
pite the fact that this involved re-loading the grain from the
granaries on to barges, and rowing or towing them up the Tiber
against the current.

Not only was land transport on a much smaller scale in classical
antiquity, but its methods and power sources also differed very
markedly from those of Western Europe before the introduction
of the internal combustion engine. The horse as a traction animal,
the mainstay of all land transport in medieval and later times,
played an insignificant part in Greek and Roman transport.
Its place was taken, for light transport by the mule, and for
heavy transport by the ox. Indeed, the whole part played by
wheded vehicles was much less, the most common methods of
transport being the human porter, or the donkey or mule with
panniers.

There are a number of very logical reasons for these preferences.
The human porter (in Latin saccarius — 'sack man') is much more
adaptable in ever)- way than a vehicle or pack animal. He can
make access for himself to the hold of a ship, can climb ladders, go
along narrow paths or gangways and is, so to speak, self-loading
and self-unloading. A very teUing piece of evidence came to light

Copyrighted material

LAND TRANSPORT

171

during the excavations at Ostia; buildings identified as granaries
had access paths and doorways which were ahnost certainly not
designed for vehicles of any kind — they are too narrow, and have
sharp bends in them. All the grain, therefore, was taken into and

out of these buildings by man-power.

The limitations of human porterage are obvious. The maximum
load, in conditions in which it had to be carried more than a short
distance (say 40-50 yards), was of the order of 50-601b (23-27 kg).
In the granaries and docks at Ostia this would mean a sack con-
taining at the most, about 4 modii (see p. 169) or, in Greek
teims, f of a medimnos. Such a load could be carried over distances
up to 3-400 yards (275-365 m.) For anything over that, it would
probably have been more practical to transfer the load to pack
animals for as much of the journey possible, and unload them
by hand at the other end. The rate of movement would be quite
slow, perhaps no more than 3mph (just under 5kph). There was
also, of course, the overriding limitation imposed by the available
supply of man-power, which became quite short during some
periods of the Roman empire.

For larger loads or greater distances the ideal mode of transport
is the mule or donkey with panniers. This was the method used by
all the small-scale transport contractors in the Greek and Roman
world, who were in fact called ^mule-drivers' (muliones in Latin,
onSlaim in Greek).

There were several reasons for preferring mules to horses. Mules
(and this normally meant the offspring of a mare and a male
donkey, not vice versa) are more amenable to the task of carrying
a load, being less temperamental than horses, and more easy to
train for that type of work. The pro\'crbial stubbornness, combined
with the pro\ erbial patience, are much easier to deal with than
the high spirits or viciousness of a wayward horse. The Roman
soldiers recognized this when they called themscl\cs 'Marius'

mules^ having been called upon, after Marius' reforms of 101 b.c,
to carry with patience a much increased weight of equipment on
route marches.* The mule's skin is harder and tougher than a
horse's, and hence less liable to damage by rubbing or chafing. It
can stand extremes of heat and cold better than a horse, and can
survive on less water for a longer period. Its hooves are harder

*Frontinu8, Strategemata IV, 7.

Copyrighted

172 ENGINEERING IN THE ANCIENT WORLD

(equines were not noimally shod), and it is much more sure-footed
on rocky paths or on the edges of steep slopes. It travels slowly but
steadily, at something over 3mph, and its slowness is partly com-
pensated by the fact that it needs only a short period of deep —

some 4-5 hours in every 24. Thus, if lightly loaded and travelling
over fairly easy ground, it can cover something like 50 miles
(80km) in a day. By a curious but typical perversity, it tends to
walk slowly down a slope and faster up it.

Pack-mules were used extensively until the early part of the
present century. From illustrations it would appear that those
used by the Greeks and Romans were comparable in size and
physique. It is therefore possible to make a rough assessment of
their performance*

The mules used by the British Army up to and during World
War I were in general between 52 and 60in high at the withers
(1.32-1 .52m or, as it was usually expressed, '13 to 15 hands').
Extra large ones were iis high as 64 in (1.62m). In weight they
ranged from about 600-900 lb (270^ 10 kg). They were reckoned
to be able to carry about 30% of their own weight as load, i.e.
about 2001b (90kg) for a small animal or 2701b (122kg) for a
large one. Naturally, if they were required to cross rough or hilly
terrain, these loads had to be reduced accordingly — perhaps by
about 25%.

There were further limitations. The load had to be of such a
shape and size as to go into a pannier or pouch suspended across
the mule's back. It dbo had to be balanced, which meant that it
had to be divisible into two roughly equ^ portions, one to be
carried on each side. For instance, even a very big mule could not
carry a block of stone weighing 2701b (122 kg); the most it could
manage 'in one lump' would be about half that weight, counter-
balanced on the other side by a similar weight.

The same considerations apply to donkeys, but the figures for
loads and body size have all to be scaled down. Their height
ranges from about 3-5 ft at the withers (0.9-1 .5m, or 'nine to
fifteen hands'). One again, ancient illustrations suggest that the
donkeys used by the Greeks and Romans were about the same size,
or perhaps rather smaller. Thus a small donkey could carry some-
thing in the region of 1201b (54kg) in its panniers, and a large
one could manage the same sort of load as a mule.

It has already been remarked that a typical small-scale transport

Copyrighted material

LAND TRANSPORT

173

contractor in the ancient world would maintain a troop of mules
or donkeys, and take on transport jobs from fanners, merchants
or anyone else who needed his services. Since his troop could be
kept employed for a lot of the time (or so he would hope), it would
be worth his while to maintain them perrnanently, growing some
forage crops if he owned any land, or renting grazing land, or even
buying fodder from local farmers. His animals would normally be
purchased from a mule stud farm. However, if a private indivi-
dual wished to carry some merchandise of his own (to market, for
instance) but was not in the habit of doing so regularly, it would
obviously be uneconomical to maintain bis own donkeys or mules
permanently. In such cases it was common practice to buy a
donkey, convey the goods to their destination and then sell the
d(Hikey, either along with the goods or else to a dealer in pack
animals, if there was one available. This is probably the signifi-
cance of the phrase in Aristophanes' Wasps (367) — *to sell the
donkey, panniers and all'.

The wheeled vehicles of the Greeks and Romans fall into two
main categories, the heavy farm wagons, normally drawn by oxen,
and the lighter \ ehicles, mainly for passenger transport, drawn by
mules, or occasionally by horses. Speaking in general terms, the
Greeks and Romans do not seem to have made any very important
advances in the design of vehicles. By contrast, the evidence from
North- West £urope, in the form of some relief illustrations from
France and Germany, and some archaeological evidence from stiD
further north, suggests that Celtic wagon-makers of tiie early
centuries ajx developed their designs to a highly sophisticated
level. It is difficult to find any convincing reasons Vfhy this should
have been so.

Almost every type of wagon in the classical world was originally
designed to be drawn by two animals, and, since the method of
attachlnt^ ihem to the vehicle was by yoke and pole, it would
seem that the ox-drawn vehicle came well before the horse-drawn.
The conformation of the ox, with its 'minor hump' at the withers,
suits this method very well, since the thrust, whether it is derived
mainly from the front or the back legs, can be taken from that
pdnt All that the harness has to do is to keep the yoke in position.
Accordingly it normally took the form of a 'throat-and-girtfa*
harness. It, a strap passing around the body just behind the frcmt
legs, which served to hold the yoke down and prevent it from

Copyrighted material

174 ENGINEERING IN THE ANCIENT WORLD

riding up over the *hump', and another passing around the base
of the animal's neck, which helped to keep the yoke from sliding
backwards, and took some of the thrust (though not much) away
from the withers (Fig. 56}. The height at which the other end of
the pole was attached to the vehicle was such that the animals
tended to push slightly upwards on the yoke rather than horizon-
tally forwards. The same was true, of course, when they were
harnessed in this way to a plough.

It is highly significant that this was also the normal method of
harnessing equines to vehicles. Here, however, the animal's con-
formadon makes it highly unsuitable. The yoke, having no *himip'
to rest on, can slide back and forth over the withers, and the har-
ness is, therefore, the actual 'power take-off' instead of the yoke.
This has all sorts of unfortunate consequences. Because a horse's
neck is longer and curves upwards more, the throat-harness tends
to ride upwards and forwards ; and since it is taking almost all the
thrust, it tends to compress both the windpipe, thus impairing the
breathing, and the surface blood-vessels, which may interfere with
the blood circulation to the brain (Fig. 57.) One attempted solu-
tion of this problem was to join the throat-strap and girth-strap
together by another strap ('martingale') between the legs. It was
eventually solved some centuries later (probably about the ninth-
tenth century a.d.) by making two fundamental changes. First,
the yoke and pole were replaced by shafts, which ran beside
the animal and lower down than the yoke, so that the pdnt of
attachment was lower. Secondly, the flexible throat-harness was
replaced by the stiff collar, which did not ride up onto the neck,

Fig. 56

Copyrighted material

LAND TRANSPORT

175

but put the pressure on the shoulders and chest of the animal. At
the same tune, the shafts ensured that the pull was equalized on
both sides.

\Ve have, therefore, the rather strange fact that the Greeks and
Romans, though in general intelligent and technologically com-
petent people, used a method of harnessing theu" horses and mules
to vehicles which was clumsy and inefhcient, because it had been
designed in the hrst place for harnessing oxen. Why should this
be so?

Fig. 57

Here is perhaps an opportune moment to look at some very wide-
ranging considerations about the whole history of science and
technology.

It has always been the case, and in all probability will always
remain the case, that applied science and technology are perma-
nendy concerned with catching up with what the society in
question sees and recognizes as its immediate requirements. These
requirements may be essential (such as improvements in food
production) or they may be luxuries (such as power-driven tooth
brushes), but in every case they represent something which» at the
time, has not yet been made practicable, and it is the task of the
technologist to make it so. In the last thirty years we have seen
perhaps the most impressive and telling example of this principle
in operation. In 1946 it was not possible to build a rocket which
would tra\'el to outer space. But two major world powers decided,
for various and complicated reasons, that there was a genuine need

Copyrighted material

176 ENGINEERING IN THE ANCIENT WORLD

for them to develop such rockets. Accordingly, the financial re-
sources were made available, and the technologists improved their
designs and increased their capabilities until the demands of the
two societies were met. This is how technology operates, and this

is what it does.

If, therefore, we encounter in an ancient society a situation in
which there seems to have been a failure on the part of technology
to meet what appears to us an obvious demand, we must ask the
following question. Was this because the technologists were incom-
petent, or was it because the demand itself never really existed?
And if we ask this question in the context of Greek and Roman
harnessing, what should the answer be ?

There is no simple or straightforward answer, but certain facts
must be put forward whidi are crucial to the question. The harness
described above was designed for oxen, to oiable them to puU a
heavy load slowly, almost certainly a plough in the first place.
Most of the donkeys and mules of the ancient world would not
have been strong enough to pull comparable loads,* and, though
there are difhcullies in interpreting the evidence, it can be said
with some certainty that their horses could not either. There is
no e\idcnce whatsoever that the Greeks and Romans had any-
thing comparable to the modern 'heavy horses' — the Clydesdale,
Suffolk Punch or Shire. As for the actual size of their horses, there
is a curious conflict in the evidence. In black-figure vase-paintings
of the sixth century B.C. they are drawn laige, compaiable in
hdght almost with a modem race-horse, with very slender legs.t
In the famous Parthenon frieze, however, the horses'in the proces-
sion are quite small. It is very difficult to decide whether this is
artistic licence — a distortion of scale to accommodate standing
men, horses and cliarioLccrs within the height of the frieze — or
whether it represents the horses accurately and life-size. However
this may be, the illustrations of working horses (which are in fact
quite rare) do not suggest that they were much bigger, if at all,
than mules. The evidence of Roman reliefs suggests that their
horses were not much bigger either.

*It is therefore very puzzling lo find that Sophocles, in his famous Ode
on the ingenuity of man {Antigone 338-41) speaks of ploughing the land
*with the offspring of horses' — usually taken to mean mules.

|e.g. Exekias' picture of Castor and Polydeuces (Arias-Hirmer-Shefton,
HisUny ofGmk Vast-Pamting, plate 63).

Copyrighted material

LAND TRANSPORT

177

The other important fact is that, where hcMrses or mules are
shown in Greek and Roman illustrations harnessed to a vehicle by
the system described above, the vehicle is in most cases quite small,
and not heavily loaded. There are a few exceptions,* but in general

the big horses are drawing chariots (for racing or warfare) and the
mules are drawing small wagons. In fact, they would be working
very well wdthin their load-pulling capacity, i.e. the load they
would be capable of pulling if efficiently harnessed. In those cir-
cumstances, the problems of the throat-and-girth harness would
not arise, or at least not in an acute form. Four highly-bred horses
might in theory have been able to pull a load of 2-3 tons at about
4-5 mph. What they did in pracdce was to pull a light chariot and
its driver — perh^ 4cwt (200kg) altogether — very much faster,
which, as we have akeady seen, was precisely ihid reason for
choosing horses. The typical Roman vehicle for fast travel (carry-
ing one or two passengers) was the cisium. Cicero, in his earliest
surviving speech, delivered in 80 b.c. {Pro Roscio Amerino
Chapter 7) speaks of a man making a night-time dash from Rome
to Ameria (about 52 miles, 85 km) in 'ten nocturnal hours'. This
(to skip the details) would mean an average speed of about 7 mph
(1 1 kph). In darkness, and on a *minor road', this was quite good
going. Cicero uses the plural (cisuf), which is usually taken to imply
a relay of vehicles and horses.

The reasons for using oxen to draw heavy loads are that they
are moderately docile (by comparison with bulls, the males which
have not been castrated) and sure-footed, and can exert a very
strcmg forward thrust on a yoke — of the order of 1^ times their
own body weight. They are, of course, very slow indeed. Under
heavy loading, they can travel no more than about 1 mile in an
hour, and if there are obstacles in the way it may take them a whole
day to co\cr a mere 5-6 miles (8-9.5 km). They have, however, a
great advantage over horses and mules in the matter of feeding.

Bovines are the best adapted of all herbivorous animals. Their
intake of food passes first into the rumen, where micro-organisms
work on it, and convert almost every type of protein present in the
food into microbial protein, which can subsequently be absorbed
into the system via the stomach and intestines and fully utilized.
Hence the colourful, but accurate description of a cow as 'a giant

♦e.g. the four-wheeled cart, laden with sacks, in the Roman relief
illustrated in RostovzefT, SEHRE« pi. XLVI, 3.

Copyrighted material

178 ENGINEERING IN THE ANCIENT WORLD

fermentation vat'. Equines, by contrast, have no rumen, and can
only assimilate certain types of protein directly. Therefore, if a
horse and an ox of the same body weight consume the same quan-
tity of the same food, the ox is likely to gain quite a lot more
nutrition from it. Some micro-oiganisms are produced in the
large intestine of equines, and they act on the (as yet) unused
protein (e.g. grass films), but after converaon, only a small amount
of the microbial protein is then absorbed through the intestinal
wall, the greater part being lost with the faeces. In fact, because
the animal has to use up its protein supply in order to manufacture
the micro-organisms, there is actually a net loss of protein over
the whole process: an interesting contradiction (apparently) of
Aristotle's dictum that 'nature does nothing without a purpose'.

This is the chemical advantage of the bovine over the equine.
There is another physical advantage, that the bovine can ingest
about 1^ times as much food as an equine. This works out at about
3% of its own body weight per day in dry weight equivalent, as
compared with 2% for a hotac For this reason, a horse cannot
survive on a diet of wheat straw alone. If it eats as much as it can
possibly manage every day, the nutrition it can extract from that
quantity is not enough to keep it alive. An ox, by contrast, can eat
1 J times the quantity, and extract or convert a higher proportion
of the protein content. This, however, represents the maximum
quantity, assuming that the fodder is freely available and the
animal eats continuously ad lib. Normally, if the diet is suitably
blended, a working animal can derive all it needs from about half
that quantity.

The Romans were well aware of these advantages (though not
of the scientific reasons for them) and found means of keeping
their working oxen on farms without having to feed them anything
expensive, or anything which could not be produced on the farm
itself. Gato recommends {De Agric, CSiapter 54) that during the
season of their hardest work (the spring ploughing, in March-
April) oxen should be fed 151b (6.8kg) of hay and 15-201b
(6.8-9 kg) of 'mash', i.e. chaff, wine-press refuse, etc. per day.
When the busy season ends, they can be fed on lupines, vetch and
other legumes, which did double duty both as forage crops and as
rotation crops to fix the nitrogen in the soil. Later in the summer
they were fed on leaves (ehn, ash, poplar, etc.) and when that
supply gave out, on hay and chaff. Gato was not given to over-

Copyrighted material

LAND TRANSPORT

179

generosity in the matter of feeding animals or slaves, but this diet
was not inadequate, even judged by modern standards.

Of the wheeled vehicles little need be said apart from some
general points. The wheels were of three types, solid, made from
planks criss-crossed, the crossbar type, and the spoked type (Fig.
58). Exactly as one would expect, the solid type ('drums', tympana)

ijb) Cross-bur (o) Spoked

Fig. 58

were used for the heaviest vehicles and the spoked typ>e for lighter
ones, with as few as four spokes on early Greek racing chariots.
The crossbar type appears rather rarely. Both two-wheeled and
four-wheeled vehicles are well illustrated. The two-wheeled heavy
vehicles are usually drawn by oxen for obvious reasons. The load
might be exactly balanced over the axle, but the chances are that
it would tend to tip one way or the other, and oxen would be better
able to cope with either a weight pushing down on the yoke or an
upward pull on the girth-strap.

Copyrighted material

180 ENGINEERING IN THE ANCIENT WORLD

One very much debated question is whether Roman four-
wheeled vehicles had fixed or movaUe front wheels. Illustrations
generally suggest that they had a fairly short whcd-base, and could
therefore be dragged around comers without much difficulty.
TTiere is no evidence to establish the use of a *swivel-table' on
which the front axle is mounted and which can be turned around
with the pole (or, on later vehicles, the shafts). It has been pointed
out that in most illustrations the wheels appear too close to the
bodv of the vehicle to allow room for them to turn. On the other
hand, since most of the illustrations are in low relief, it might be
aigued that the artist was prevented by his medium from indica-
ting a gap between wheel and body. There is also one scrap of
inscriptional evidence in the so-called *price edict' of Diocletian,
the text of a decree fixing maximum prices for a very wide range
of goods, issued in 301 A.D. Among parts of wagons there appears
the item columella, a little pillar*. Hiis is the word used for the
short spigot on which the rotor of an olive-press turned, and, as it
happens, every other use of the word relates to a vertical spigot or
post. It would therefore be most natural to assume that in this

context it refers to the stub or spigot on which a 'swivel-table'
turned.

Another minor question is how the wheels were mounted on the
axles, since it is impossible to tell this from illustrations with any
certainty. On lighter vehicles it is probable that the axle was fixed,
and the wheels turned on a short 'stub' at each end, being prevented
from coming off by a 'lynch-pin' passing through the stub (Fig.
59a). To avdd wear on the pin and tiie wheel-hub, there was
normally a metal washer between them. On heavier vehicles it is
more probable that the wheels were fixed to the axle, which turned
round in some sort of bearing on the under-side of the chassis
(Fig. 59b). Both these methods of mounting are in use today —
the fixed stub on the non-drive axle of a car, and the rotating axle
on railway trucks. The important dilTerence is that wear shows
itself in different ways. As soon as the wheel gets slightly loose on
the hxed stub, it begins to tilt one way or the other, which means
that the area of contact is reduced, and the wear is concentrated
at one end of the hub at the top, and the opposite end below (Fig.
60). With the revolving axle, however, the wear takes the form of
a groove in the axle where it rubs on the bearings, but does not
cause tilting of the wheels, and is therefore safer if the vehicle has

Copyriyi uuj I . id^Li lai

LAND TRANSPORT

181

Fig. 59

to be taken over uneven ground. Virgil speaks of the 'screeching
axle' of a heavy cart, and he may well be stating the literal truth.

There is very little evidence for the use of lubricants in antiquity,
except on racing vehicles, when the friction might be so great as to
give rise to a fire risk. Water seems to have been the most commonly
used, but it cannot have been very effective.*

♦See H. A. Harris, Lubrication in Antiquity, Greece and Rome XXI/1,
April 1974, 32-36. Prof. Harris was probably right to express scepticism
over the supposed 'roller bearing' of the DE JBJERG wagon (Oxford
History of Technology, Vol. II, p. 551). The holes might well have been
drilled in order to remove the axle-socket, and left there to contain
lubricant — perhaps grease.

Copyriyhica niaiciial

182

ENGINEERING IN THE ANCIENT WORLD

Wheeled vehicles were of course normally driven on roads or
farm-tracks, but there is evidence that in a few places in the Greek
world special 'tramways' were built of stone slabs, with deep
grooves for the whceb to run in. It is surdy not a ooinddenoe that
the loads being carried were in each case abnormally heavy —
blocks of stone from a quarry or, in one famous case, warships.

When remains of such a 'tramway' are excavated, there is
naturally some doubt as to whether they were deliberately con-
structed as ^tramways' or whether they were ordinary roads in

which grooves have been worn by the passage of many vehicles
over a long period; once a shallow groove had been formed, the
wheels of later vehicles would naturally tend to slide into it, and
follow it. There are, however, certain features which suggest that
the grooves were deliberately cut. They have a remarkably con-
sistent width and depth (20-22 cm and 12-15 cm respectively) and
there seems to have been a 'standard gauge' of 11 2-144 cm.
But the most telling piece of evidence is that the grooves are
often found to pass across the centres of the stone dabs. It

Hub

Fig. 60

Copyrighted material

LAND TRANSPORT

183

would be a remarkable coincidence if this had happened by
accident.

Much the most famous of these 'tramways', the Diolkos, or 'pull-
across', ran across the Isthmus of Corinth from the Corinthian
gulf to the Saronic gulf, and was used by the Corinthian navy to
transfer ships back and forth.* It followed the shortest route
possible, avoiding steep gradients, and ran not very far from the
modem Corinth canal. It is not certain what motive power was
used. Because there are no obvious marks of animals* hooves on
the roadway, it has been argued that the ships, mounted <m
wheeled trolleys, must have been manhandled across, probably by
their own crews. This argument is plausible, but not conclusive.

In other situations it was necessary to transport abnormally
heavy loads overland without using wheeled vehicles at all —
where a stone tramway was impracticable or, because it would
only be used for a short time during a particular construction job,
uneconomical to build. Vitruvius describes ways in which column
drums and architraves were transported (X, 11—14). One
ingenious method was to use a column shaft or column drum
(which would be roughed out round at the quarry) as a roller, by
constructing a wooden frame around it, and fixing short iron
spigots in each end (by making a socket and pouring in lead) and
fixing bearings in the wood frame in which the spigots turned
(Fig. 61a). The frame was then harnessed to a team of oxen, and
pulled along like a heavy road-roller. This device was invented by
Chcrsiphron, architect of the temple of Artemis (Diana) at
Ephesus.

His son Metagenes showed similar resourcefulness. The archi-
trave blocks were of comparable weight but square or rectangular
in cross-section, so they could not be rolled along. Accordingly,
two wheels were made about 12ft (3.66m) in diameter, probably
of the cross-bar type (Fig. 61b), with heavy bars and broad
felloes, so they did not easily sink into the ground. The ends of the
architrave blocks were ^enclosed' in the wheels (probably between
the cross-bars) and fitted with a spigot at each end whidi, as with
the column drum, turned in a bearing mounted in a wooden
frame. The spigots would have to be lined up with the centre

•See B. Ashinole, Architect and Sculptor in Classical Greece (Phaidon
Press 1972), pp. 20-21.

Copyrighted material

184 ENGINEERING IN THE ANCIENT WORLD

of gravity of the block to allow the wheels to turn steadily. Vitru-
vius stresses that the whole operation depended on two conditions,
that the quarries were not far away from the building site (7^
miles, 12 km) and the ground in between was flat and level.

Another engineer in Vitnivius' own day, Paconius, tried to go
one better, and paid the penalty for his ambition. He had to convey
a block of stone 12ft X 8ft X 6ft (to replace a statue base) from

Copy My I uuu 1 1 idiLi icil

LAND TRANSPORT

185

the same quarries over the same distance. This volume of sume
(576cuft9 16.32m*) would have weighed something in the region
of 45 tons (45.7 tonnes). Paconius used the same method of motmt-
ing the block in two wheels 15ft (4.57m) in diameter, but the

method of traction was different. He fixed battens of wood two
inches (5 cm) thick (which sounds incredibly slender) from wheel to
wheel 'in a circle around the stone, at a distance of less than a foot
apart'. Vitruvius does not say whether these were on the rims of the
wheels or in towards the centre. Paconius then wound a rope
around this 'bobbin', and hitched it to the team of oxen. As they
pulled, the rope unwound and caused the 'bobbin' to revolve, thus
rolling the wheels along the road (Fig. 61c). This is quite a sound
method, as it gives a slight mechanical advantage; unf<nrtu-
nately, however, there was a snag. Oxen harnessed to a frame by
the older method were aUe to steer the wheels along the road, but
Paconius' structure could not be controlled in that way, and kept
swerving to one side or the other. It had then to be pulled back
on to the track again. In the end the job took so long tliat Paconius
ran out of fodder for his oxen, and went bankrupt.

This chapter began by contrasting the speed and effectiveness
of sea transport with the slowness and clumsiness of land transport
in the classical world. Two Roman poets expressed this contrast
very aptly in two short poems, the second being a direct and witty
parody of the first GatuUus wrote (poem IV) of the small vessel
which had brought him home to Italy from Asia Minor:

My friends, that little yacht you see over there
Claims to have been the fastest vessel in the world;
There was no keel afloat j cutting through the water.
But she could overhaul it, whether called upon to fly
With oars a'Splashing or with linen sails aloft . . .

Among the works attributed to Virgil's early years is a short poem
about a ti ansport contractor in the area of his birthplace near
Mantua;

My friends, that man Sabinusyou see over there
Claims to have been the fastest mule-man in the world;
There was no racing curricle, hurtling along the road
But he could overtake it, whether called upon to fly
Post-haste to Town, or down the muddy lane to Brixia . . .

{Catalepton X)

Copyrighted material

8

The progress of theoretical knowledge

Since the Roman contribution to technology, though consider-
able, was ahuost entirely in the field of practical application, the
state of Greek theoretical knowledge may be regarded, for all
intents and purposes, as that of the whole Mediterranean world
and the Roman Empire, down to the fifth century a.d. or even
later. It is therefore unnecessary (and would in any case be far
beyond the scope of this chapter) to review the subject as a whole.
It will suffice to highlight one characteristic feature of Greek
thought, and then to assess the state of theoretical knowledge in
three particular areas which are closely related to the main topics
dealt with elsewhere in this book — hydrostatics, mechanics and
chemistry.

It is possible to see, in almost every branch of Greek literature,
a particular trait of the Greek mind which had important effects
in some branches of scientific thought. It was a liking for stability,
rest and permanence, and a corresponding dislike, almost a mi»-
trust, of change, movement and Ytkat they called genesis and
phthoray *coming-to-be' and *passing-away'. Why this should be
so is something of a mystery, but perhaps their very acute aware-
ness of the impermanence of physical things in their world, and of
human life itself, caused them to set a high value on the permanent
and the stable. However that may be, one result was that their
understanding of static conditions (e.g. hydrostatics, or mechanical
problems not involving movement) was very acute, wheieas their
ideas on dynamics and ballistics were surpriangly incomjdete and
inaccwate. They spoke of velocities, relative velocities and reast-
ance, but hardly even began to study acceleration or deceleration,
and they had only a rather vague notion of inertia or kinetic
energ)\ They observed that a stone continued to fly through the
air after it had left the hand of the thrower, but throughout

Copyrighted material

PROGRESS OF THEORETICAL KNOWLEDGE 187

antiquity they ccmtinued to offer some quite absurd eacplanatioiis
of why it should do so.

When the Greek philosophers first began to enquire into the

physical nature of the universe, it was prcciscl)^ the problem of
change and of 'coming-to-be and passing-away' that preoccupied
them and, perhaps, disturbed them. More than a century later
Plato was still bothered by the same problems, and though his
thought was untypical in several important respects, on this point
it accorded completely with the Greek tradidon. He went so far
as to say that physical objects, because they undergo perpetual
movement, change and destruction, cannot be 'known' or 'under-
stood' in the true sense of those terms, and one of the aims of his
famous theory was to supply, in the 'Forms' or 'Ideas', eternal
and unchanging objects of true knowledge, by relation to which
material thmgs could be studied and reasonably interpreted,
though never truly 'known'. One consequence was what might be
termed an 'amti-physical' trend in Plato. Though his Timaeus is
devoted to a detailed (if rather curious) analysis of the physical
world and its creation, he elsewhere exhorts the true philosopher
to turn his back on such matters, and 'rise above them' in the
pursuit of true wisdom and knowledge. Plato's admirers see this
as the honest logical conclusion hrom his theory of Forms and his
epistemology. His enemies see it as the snobb^ contempt of an
aristocrat and man of independent means for those of his fellows
who had to deal with physical objects all their lives, and worked
for their living.

This feature of Plato's thought would be less important were
it not for the fact that some of his followers in later centuries not
only accepted the 'anti-physical' attitude, but carried it even
further than Plato had thought fit. Uncritically, and without
really understanding his arguments, they exalted the 'pure' and
theoretical sciences (such as geometry and astronomy) and looked
down on any research that was mechanical, or which had practical
applications. Accordingly, if they admired a particular scientist
or thuiker, they tended as a matter of course to attribute to him
all their own prejudices. Plutarch, writing in the first century a.d.,
is guilty of tlds in speakmg of Archimedes, and the much-quoted
passage in which he does so (MarceUus 17, 3-^) deserves to be
treated with the utmost scepticism.

There were other reasons for the lack of progress in the study of

Copyrighted material

188 ENGINEERING IN THE ANCIENT WORLD

motion. Much the most important of these was the ahsence of
devices which could be used to measure short intervals of time, of
the order of a few seconds. They had water-docks of various typesy
some of them capable of running for up to 24 hours or more, but

when they were used for scientific measurements (e.g. in astro-
nomy) they did not measure the passage of time as such, nor were
they calibrated in time units. They measured one time interval as
a multiple or fraction of another, such as the time between the
first appearance of the sun at dawn and its complete clearance
above the horizon, as a fraction of the whole diurnal cycle. Yet
another obstacle to research was their method of dividing that
cycle into 'hours*. Instead of having 24 horns of equal length, they
divided the day, measured from sunrise to sunset, into 12 equal
divisions, with the result that their 'hours' were not fixed imits of
time, and varied from what we would call *about 40 minutes' in
winter to 'about 80 minutes' in summer. This system was adequate
and indeed rather useful for the time-measuring needs of everyday
domestic life. The working day for the ancients ended effectively
at sunset, and, if one had to get through one's daily duties in less
time in the winter, it was really just as sensible to have shorter
hours as to have fewer hours. But for the astronomer it created
serious problems. If he wished to fix a point of time at which an
eclipse had been observed, he had to specify the 'hour' and also
the exact time of year, and do quite a lot of arithmetic to correlate
two points in time fixed by this method. It was probably for this
reason more than any other that our system ci fixed equinoctial
hours eventually replaced the 'classical' system. Until that occur-
red, it was virtually impossible to subdivide the hour into minutes
and seconds, which would likewise have varied in length accord-
ing to the time of year — a truly daimting idea.

There was, however, one smaller unit of measurement, used in
the law-courts of Greece and Rome, which was to some extent
fixed and regulated by law and did not vary with the seasons. In
order to ensure fairness, the prosecution and defence were each
allowed the same amount of time for the presentation of their
cases. The measurement was made by the simplest form of water-
clock, called a clepsydra* It was a jar of fixed capacity with a hole
at its base of fixed diameter, through which the water ran away.
The time was assessed in terms of 'two clepsydras' or 'five clepsy-
dras', and so on, according to the complications of the case or, in

Copy I iyl uuU I Mdj-I lal

PROGRESS OF THEORETICAL KNOWLEDGE 189

a dvil suit, the amount of money at stake. There may have been
some attempt to standardize this unit of time, since the surviving
remains of one clepsydra (in the Agora Museum at Athens) bear

two marks which seem to indicate a clieck on its capacity by an
inspector of weights and measures. A rcconsti ucted model of this
clock runs for about 6 minutes f>er filling.

But this use of a time-measuring device was in the interests of
fairness, not in the pursuit of science. The same might be said of
another application of the same device, recorded in an anecdote
concerning a well-known prostitute in fourth-century Athens, who
was known professionally as 'The Clepsydra* because, in order to
ensure a fair distribution of her favours, she installed a water-clock
in her boudoir, and timed her clients' visits by it.

The science of hydrostatics was given its theoredcal basis by
Archimedes, in the latter half of the third century b.c., as set out
in his surviving work, Peri Ochoumendn — *On Floating Bodies'.
Of course, a good deal of practical application had been going on
for many years on such things as siphons, water-clocks and buoy-
ancy devices, but Archimedes codified the theory, and gave it a
sound mathematical basis.

The treatise beg^ with a basic supposition about the nature of
liquids, which deserves quotadon in full: 'Let it be assumed (i.e.
accepted without proof or demonstration) that the nature of liquid
is such that partictes on any one level will (fisplace (literally, **shove
aside") other particles which are on the same levd but under less
pressure : and that the pressure is determined by the liquid which
is perpendicularly above each particle, provided that the liquid
is not in an enclosed container, and under pressure from some other
force (i.e. other than gravity).'

He goes on immediately to make clear that by 'on the same
level' he means 'at the same distance from the earth's centre',

acceptmg ^thout comment the conclusion reached by fourth-

century astronomers that the earth is spherical, and that the force
of gravity acts towards its centre. These ideas, so long familiar to
us, cannot have been obvious to common sense, easy to assimi-
late. 'N^rtuaUy all the rest of the treatise consists of deductions
from these basic assumpdons, the favourite type of argument being
the reducHo ad ahsurdum — 'if this proposition is imtrue, we shall
be led to a concltision which is contrary to oiu* original assump-
tion'.

Copyrighted material

190 ENGINEERING IN THE ANCIENT WORLD

He first shows that the surface of any mass of liquid in a stable
dispodtion is a sphere, or segment of a sphere, concentric with the
earth. The proof is simple. If the surface is anything other than

spherical and concentric with the earth, there will be particles on
the same level which are under different pressures, and movement
will result. So the disposition will not be stable. In contrast to this
straightforward logic, there is evidence, recorded by Pliny (Nat.
Hist. 2, 65, 164), that the question was tackled empirically.
The curvature of such a surface is, of course, not discernible
except over an expanse of several miles, but Pliny points out that
a shining object at the top of a ship's mast remains visible from the
shore after the hull has disappeared below the horizon.

Archimedes then (I, 3) passes on to deal with three categories of
body, those whose density is (a) equal to, (b) less than, (c) greater
than, that of the liquid in which they are immersed. A body less
dense than the liquid will float, and the portion of it which is sub-
merged bears the same ratio to the whole volume of the body as
its whole weight bears to that of an equal volume of the Uquid. If
it is forcibly held bdow the surface it will exert an upward thrust
equal to the difference between its own weight and that of the
liquid it displaces. This thrust was used by Gtesibius to work an
automatic water-level control in the water-clock (discussed below),
and it fufils the same function in the ball-cock of a modern water
cistern. A body whose density is greater than that of the licjuid
in which it is immersed will sink to the bottom and, if weighed
while submerged, will appear lighter than its true weight by an
amount equal to that of the liquid it displaces.

In addition to these theoretical principles, we know from other
soiu'ces that Archimedes devised a practical method of using them
to assess the proportions of gold and silver in a crown made for the
king of Syracuse. To do so, it was necessary to measure the exact
volume of the crown, and his discovery of a method for doing this
must surely be the most famous story in the history of science. On
stepping into an over-filled bath-tub, he saw that the water which
overflowed, if caught and measured, would gi\ e the exact volume
of an irregularly shaped body — namely, his own. In his haste to
get home from the public baths and try this out on the crown, he
made his well-known 'nude dash' through the streets of Syracuse,
with shouts of ^Hiurekay Heureka' (1 have found it'). This

Copyrighted material

PROGRESS OF THEORETICAL KNOWLEDGE 191

must have mystified the onlookers, who probably were thinkmg
that exactly the opposite had occurred.

Next m the treatise comes a series of problems m stability — the

tendency of a floating body to assume a particular position, and to
return to that position if tilted away from it and then released.
Apart from two propositions on segments of a sphere (I, 8 and 9)
they all deal with one particular shape — the paraboloid (the solid
traced out by a parabola, which Archimedes calls a 'section of a
right cone', revolving on its axis — Fig. 62). Its stability is assessed

Axis

Fig. 62

in relation to two factors, (a) its shape — the proportion of its
height to its breadth, which is measured in a very sophisticated
way — and (b) its density or, as we would say, its specific gravity.
Why Archimedes should have been so preoccupied with this parti-
cular shape 18 not clear; but two explanations have been offered.
One is that seen in cross-section it approximates very roughly to
the shape of an ancient ship's hull, and that he was thinking, in
the very long term, of calculations ^N/bidL might be useful to a naval
architect. The more likely alternative is that, having wresded with
various problems connected with the parabola, and having
brilliantly solved two of them (measuring the area of a segment
and finding its centre of gravity) he felt a kind of affection for this
type of curve, and for the solid which it generates.

Copyrighted material

192 ENGINEERING IN THE ANCIENT WORLD

In contrast to this type of method, we have a few remains of
hydrostatic theory in which arguments of a kind which might be
called empirical are used. They are preserved in the introducdon
to Hero's Pneumatica, and they include one very striking example
of a truly scientific method. Though this is a matter of controversy,
there is no real proof that the idea for the experiment was not
Hero's own.

(a) Normal siphon

Inner
Imb"

OuUrUmtf

How,

according
totfieory

r\

in prcLctico

n

Fig. 63

He has just given an eiqilanadon of why a liquid £k>W8 through
a siphon, which he presumably found in some earlier writer, and
which goes like this. If the pipe is of uniform diameter there is a
greater volume, and therefore a greater weight of water in the
longer (outer) limb (Fig. 63a). This 'overpowers' the lesser weight
in the inner limb, and draws it up (on the analogy of a pair of
scales). Since air cannot get into the inner limb, and a continuous
vacuum cannot be formed, liquid from the vessel must flow in to
replace that which is drawn off, and so on. 'But', says Hero, *we
can demonstrate that this explanation is wrong. Construct a siphon
with the inner limb long and slender, and the outer limb shorter
and much thicker, so as to contain a greater volume of liquid than
the inner' (Fig. 63b). According to the theory, the greater weight
ought to fall, and raise the lesser weight in the mner limb. Then

PROGRESS OF THEORETICAL KNOWLEDGE 193

comes the aiidal sentence: 'But tins does not actually happen;
therefore the esqilanation I have quoted is not the true one'. All
the elements of modem method are here; the formulation of a

theory, the use of an experiment (with specially designed appa-
ratus) to test it and, most important of all, the acceptance of the
experiment as a conclusive proof that the theory is wrong. (What
actually happens is, of course, that the liquid flows 'the wrong
way' thi ough the siphon.) There are very few examples of experi-
mental method in Greek science, but it is not true to say that there
are none at all.

Both the Archimedean logic and Hero's more empirical
approach do, however, share one feature in common — the Greek
trait discussed at the beginning of this chapter. On Archimedes'
basic assumption, if a liquid is in an unstable disposition (particles
on the same level under different pressures), movement will take
place and will continue until a stable disposition is reached; but
he never, apparently, tried to analyse that movement, or to assess
its speed or changes in its speed. He was interested only in the
situation of potential movement before it started, and that of zero
movement after it had ceased — in other words, the static aspects,
not the dynamic. Hero goes a littie way in that direction, but not
far. He describes the conditions under which a siphon will run,
and those under which it will stop running, and he also points out
that the rate of flow through it is not constant, but diminishes as
the system approaches the *stop' conditions. But he makes no
attempt to work out a mathematical formula for this change in
the rate of flow. It had caused his predecessor Gteabius certain
problems in connection with water-clocks. The diminution in the
rate of flow also occurs, as Hero points out explicitly, in 'a vessel
with a hole in the bottom', which must mean a clepsydra. As the
water level falls, the rate of flow through the spout slows down,
with the result that one cannot measure fractions of the time imit,

such as 'half a clepsydra' or 'one-eighth of a clepsydra', by filling
the jar half full, or one-eighth full, as the case may be.

C^esibius* solution of the problem did not involve tackling the
dynamics of the system. He merely stabilized the rate of flow and
made it constant — in fact, he took the dynamics out of it. Instead
of measuring the fall of level in an emptying jar, he kept it constant
with an automatic valve worked by a float, similar to the needle-
vahre in the float chamber of a modem carburettor (Fig. 64a).

Copyrighted material

194 ENGINEERING IN THE ANCIENT WORLD

This constant outflow was used to fill a straight-sided vessel, and
the rise of level in that vessel could be measured, and subdivided
mto fractions as required. Hero's solution, on the same basis, was
to construct a siphon with a float on its inner limb, and a mounting
which enabled die siphon as a whole to rise and fall with fluctua-
tions in the water level (Fig. 64b). As he pcnnts out, the rate of flow
depends on the 'head' of water — the difTerence of level between
the surface of the liquid being drawn oflf and the outlet of the
siphon which, by this arrangement, is kept the same at all times.

Fig. 64

Our information on Greek theoretical knowledge of mechanics
comes almost entirely from two works, Mechanical Problems,

attributed to Aristotle but probably of later date, written by a
member of liis school, and Hero's Mechanica, discussed in Chapter
9.

Mechanical Problems cover some of the same ground as Hero's
Mechanica — pulley systems, the wedge and the lever. At the start,
the author plays about with certain rather paradoxical thoughts
about circles. A typical example is that if a radius revolves, the
points along it move together and in a line with each other, and
yet those further out from the centre cover a greater distance than

Copyrighted material

PROGRESS OF THEORETICAL KNOWLEDGE 195

those nearer the centre. This leads onto a number of principles of
levers, which are concdved in the 'classic' form of mover, force,
lever, fulcrum and load (Fig. 65a). The longer the lever (between
mover and fulcrum), the greater the force exerted on the load, and

the weight of the load and the force needed to lift it are inversely
propordonal to their distances from the fulcrum (Fig. 65b), and
soon.

Fig. 65

A number of attempts are made to use this as a model to inter-
pret various mechanical systems, several of yAdck involve ships
(probs. 4r-7), The oar is seen as a lever, the rowlock as a fulcrum
and, quite correctiy, the sea as the *load' which is moved. More-
over, this movement of the load causes a 'counter-thrust' on the
fulcrum which propels the ship forwards. When we come to the
rudder (prob. 5) the model does not work so well. The point at
which the rudder is fixed to the ship is the fulcrum, the whole
rudder is the lever, the sea is the load and (here he goes astray) the
helmsman is the moving force. Once again, because the *load'
is thrust in one direction by the 'lever', the fulcrum (and with
it the stem of the boat) is thrust in the opposite direction. The
writer is groping towards the Newtonian principle that 'action
and reaction are equal and opposite' without actually arriving at
a formulation.

Another of the ship problms (no. 4) is a telling example of the
static conditions being understood, but the dynamics viewed

Copyrighted material

196 ENGINEERING IN THE ANCIENT WORLD

wrongly. It is stated that the towers amidships contribute mort to
the propulsion of the ship than those near the bows or stem,
because they sit further inboard, and have a longer lever' (i.e. the

oar between the hands and the rowlock). They therefore exert a

greater force on the *load', and though this is quite true, the fact
that their oar-blades move more slowly is ignored. The lever model
breaks down completely in prob. 6, where it is used to account for
the fact that if the mast, yard and sails are higher, the ship will sail
faster.

Problems 8, 9, 10 and 1 1 deal with wheels and rollers, and show
some understanding of the problems of friction which arise in
designing carts and wagons, but the writer's preoccupation with
drcles and leverage leads hun sometimes to omit the obvious —
e.g. that a laige wheel turns more easily than a small one because
it does not go so far down into the potholes.

The nearest approach to a concept of kinedc energy comes in
prob. 19. Why is it that, if one places an axe on a log of wood, and
a heavy weight on top of the axe, it will not spilt the wood, yet, if
the axe is swung dowTi onto the log it will split it, even though the
axe is lighter than the weight which was placed on it? 'Is it,' he
asks rather tentatively, 'because all work is done by means of
movement, and one weight is capable of imparting more move-
ment to another if it is moving than if it is stationary ?' Once again,
an important concept is hinted at but not explored in detail.

The last four problems deal with motion, and show the author's
characteristic Greek shortcomings. Indeed, in prob. 32 (if the text
is correctly restored) he seems to suggest that there is no point in
exploring such problems, because tiiey are insoluble. From the
variety of alternative e3q)lanations he gives for the deceleration of
a missile, it is clear that there was no single agreed theory. In the
final problem, however (no. 35) his preoccupation with the proper-
ties of circles leads to a fundamentally correct explanation of a
puzzling phenomenon.

Several of the early philosophers had suggested that the universe
was created from an original state of chaos in which all kinds of
matter were mixed up at random, by a circular motion, or ^eddy*.
This motion caused the lightest partides (those of a transparent,
fiery substance called aether) to fly outwards to the periphery and
form the sky. The next heavier kind, those of air, took up a lower
position, those of water came next, and the heaviest of all, those

PROGRESS OF THEORETICAL KNOWLEDGE 197

of earth, came together at the centre. At first sight, a modem
scientist might take tliis as curious reversal of the truth, since the
heavier particles, having greater weight, and thus subject to a
greater centrifugal force, would be more likely to fly out to the
periphery than the lighter ones. But it must be remembered that
the Greeks had no means of investigating the behaviour of par-
ticles moving in a vacuum. They thought of the 'eddy' as a move-
ment of all matter, mostly air, which carried the heavier particles
around with it. The illustrative model they had in mind was, in
fact, a very homely one — a bowl of soup with lumps in it of
various sizes and weights. If set in motion it would, as it *eddied'
around the bowl, exhibit exactly the behaviour which the philo-
sophers assigned to the universe — the Inggest and heaviest lumps
(of turnip, perhaps) would gather at the centre first, and the lighter
ones towards the outside. This is how (at the second attempt) our
author explains what is happening. 'Or is it because those particles
which the eddying water cannot control, because they are too
bulky and heavy, must inevitably be left behind by it, and travel
more slowly?'. He goes on to explain that if they move inwards to
a smaller orbit they will in fact be travelling more slowly, although
in a sense all the 'circles' aic revolving at the same speed (as we
would say, their angular velocity is the same). The result is that
they keep on movii^; in to a smalller orbit until they reach the
centre.

The Greek achievement in chemistry presents us with an odd
contrast. On the one hand, Greek doctors knew and used a very
wide range of chemicals, both mineral substances and others ex-
tracted from animals and plants by various processes. But on the

other, there was virtually no general theory about chemicals, or
the laws which govern their combination or separation. The
explanation for this must lie in the fact that the two types of know-
ledge are acquired in very cliircicnt ways. Tlie ancient doctor
could try out any one of the collection of chemicals in his cupboard
(from which, of course, the known harmful poisons had been
excluded) and see if it helped his patient's condition. If it did, he
could then prescribe it for other patients with similar complaints,
and if it did not, he could try another. (This procedure is rare in
modem medical practice, though not altogether imknown.) But
from start to finish, he did not reaUy need to know the chemical
content of his medicines, or to understand the real reasons why

Copyriyhioa inaici lal

198 ENGINEERING IN THE ANCIENT WORLD

they were beneficial in some cases. All he needed to know was the
source of the material, the method of preparing it and (dosdy
related to that) the dosage. One striking example illustrates how
far this empirical method could take them, and how the lack of
techniques of chemical analysis and testing prevented them from
going any further. Dioscorides (first century a.d.) the author of the
best knowTi ancient work on Materia Aledica, mentions a drug
prepared from mandr agora officinalis which, he says, can be used
as an anaesthetic. We have no other evidence for its being so used,
probably because there was no means of testing the strength of a
dose, which had to be accurate within very narrow limits.

In fact, chemical testing and analysis was, for the Greeks,
almost entirely a matter of using the senses. Of the range of terms
they used for what we would now call chemical properties, there
is hardly one which does not relate to taste, smell or touch. When
they speak of unripe fruit being 'sharp' {oxy) it is tempting to
translate it as 'add^ but if we do, we must forget all about litmus
paper or any other chemical reaction. All the Greek word means
is *something which tastes like vinegar'. Incidentally, there was
no word for alkali, simply because they did not regard alkaline
substances as being linked by a common taste factor.

Their understanding of the chemical effects of heat was also
imperfect. Because a number of familiar chemical processes are
speeded up by an increase in temperature, they tended to think
that the heat was itself the cause of the process, and not merely
one of the controlling factors.

However, when all this has been said in disparagement of their
chemical theory, it is as well to recall that the Greeks and Romans
managed their practical chemistry with fair success. They could
cure some diseases; they could brew palatable and potent wines;
they could devise an adequate and healthy diet from terribly
limited resources of food production ; and they could put a black
glaze on their pottery which still looks shiny and new today and —
who knows — may still remain so when our own civilization has
passed into history.

Copyrighted material

9

The principal Greek and Roman writers
on technological subjects

Some of the most difficult problems in the study of Greek and
Roman technology arise from the fact that our information comes
itosa a wide variety of sources. Odd references to such subjects as
metallurgy, transport or building technology are scattered about
in the works of many authors, and the archaeological evidence
comes from a large number of sites, and ranges over a long period
of time.

A complete review even of literary sources, therefore, would be
a very lengthy and tedious undertaking, but there are three ancient
writers, Hero of Alexandria, Vitruvius and Frontinus, whose works
(or some of them at least) are wholly concerned with technological
subjects, and these writers will be considered in some detail.
A fourth, Pliny the Elder, compiled a great encyclopaedia in
which technological subjects are referred to (along with almost
every other subject under the sun), so he deserves at least a brief
mention*

H£RO OF ALEXANDRIA

Hero (the Greek spelling of his name was Heron, but the Latinized

form is more commonly used) is one of the most important sources
of our knowledge about ancient technology. His works have sur-
vived in some quantity (though they are not complete), and he
was a very versatile man, with the result that he supplies informa-
tion on pure mathematics, physics, mechanics, conjuror's appa-
ratus, surveying instruments and many other items plus, most
interesting of all, some occasional insights into practical engineer-
ing at the 'nuts-and-bolts' level. A number of his devices have
already been discussed in the appropriate chapters.

Of die man himself we know virtually nothing apart from what
can be inferred fixnn his writings. His name suggests that he Hved

Copyrighted material

200 ENGINEERING IN THE ANCIENT WORLD

and worked at Alexandria; several later writers also referred to
him as *Hero the machine-man' [mechanikos). Q£ his social status
we know nothing, except that he was an educated man, well-read
in the works of the mathematicians and engineers, particularly
those of Citesibius, the most famous in the ancient world. Hero's
position, however, was probably very different from that of
Ctesibius, who Hved in the third century b.c. when Egypt was
ruled by the Ptolemies, hereditary kings descended from one of
Alexander the Great's Macedonian generals. They were extremely
wealthy and are known to have patronized both literature and
science. A number of surviving technical works are addressed
(either as a courtesy to a present patron or in the hope of attracting
new patronage) to them and to other contemporary Hellenistic
rulers, who could offer similar, though perhaps less lavish, benefits.
For instance, Biton's work on Engines of War and CatapuUs was
written for King Attains of Pergamon. But if Hero lived in the
first century a.d., as will be argued in the next paragraph, the
Alexandria of his day was under Roman rule. Cleopatra, the last
of the Ptolemies, came to her much-dramatized end in 30 B.C.,
and Egypt thereafter was kept under very strict control by the
Romans, being one of their most valuable sources of imported
wheat. On the other hand, there is nothing in Hero's writings to
suggest that he worked for the Roman government or for a Roman
patron, as many of his Greek contemporaries did. He sometimes
uses Roman names for weights and measures transliterated into
Greek, but that was common in legal documents and business
agreements made in Egypt under Roman rule. He also mentions
Latin equivalents for Greek technical terms, which seems to suggest
that he was acquainted with some Latin technical literature (which
is surprising, and that he h(^>ed to find some readership among
those Roman engineers who could read Greek.

The question of his date has been much disputed, suggestions
ranging from the second century B.C. to the second a.d. or even
later. The fact that he recognized the existence of the Romans at
all would seem to rule out a date earlier than mid first century B.C.,
and makes a later date much more probable. The question was
very open until in 1938 the machinery of modern science was
brought in. In the Dioptra Hero describes a method of calculating
the Great Circle distance from Rome to Alexandria by observing
the same eclipse of the moon in both places, and measuring the

Cc, , y od material

PRINCIPAL GREEK AND ROMAN WRITERS 201

time interval between (Chapter 35). Though he seems to be talking
in hypothetical terms (let us suppose that an eclipse has been
observed . . he chooses a date 10 days before the spring equinox.
It is surely not a coinddence that between 200 B.a and a.d. 300

the only eclipse which met the required conditions (visible both
from Rome and from Alexandria) occurred, according to modern
astronomers, on March 13 a,d. 62. Though this does not prove (as
has been asserted) that Hero was alive then and recorded the event
in his Dioptruy it does show that he knew of an attempt to calculate
the distance which was made on that occasion, and that his
Di Optra must date from later than a.d, 62. Since the geographer
Claudius Ptolemaeus, writing about the middle of the second
century A.D., uses slightly more sophisticated methods, we may
safely assume that Hero was active (a not inappropriate term)
between about a.d. 50 and aj>. 120. This is also consistent with
the probable date of his CkeirobalUstra, discussed in Chapter 5.
It was the age of the Flavian emperors at Rome, V espasian, Titus
and Domitian, and their successors Nerva and Trajan.

Hero's surviving works are here listed in the order of the stan-
dard edition in the Teubner Libra^^

>

(1) Pneumatica. There is no really satisfactory English transla-
tion for this title. 'PTieumatics' will hardly do. Some of the devices
described do use compressed air, but others use siphons and steam
pressure. The German *Dnickwerke' translates it exactly, so per-
haps 'jpfcssure-mechanisms', if clumsy, is the nearest one can get.

The work, in two books, begins with a theoretical discussion of
the properties of air, the impossiHlity of a continuous vacuum,
and the behaviour of liquids acted upon by gravity. Hero is clearly
dependent on earlier writers for this section, but his originality
consists in drawing together two schools of thought with very
different styles and interests. The first part is derived from Strato
of Lampsacus, who was the third principal of Aristotle's school,
the Lyceum, from about 288-268 b.c. In keeping with the Aristo-
telian tradition, he tends to be descriptive without quantification
and hence almost completely without mathematics, and he also
uses empirical arguments and demonstrations to explain or corrob-
orate his theories. But in the second part of the introduction, and
in Qiapters 1 and 2 where he is discussing siphons. Hero draws on
Archimedes, who started, not from emphrically demonstrable ideas,
but from certain assumptions (for which he offers no proof) about

202 ENGINEERING IN THE ANCIENT WORLD

the nature of liquids. His deductions are all strictly quantified, and
in some cases highly mathematical.

After the introductory matter, Hero describes some 75 structures
or devices, the great majority of which are ornamental and enter-
taining rather than useful. Many of them are for dispensing wine,
or mixing wine and water, including a trick jug (I, 9) with two
concealed air inlets, from which one can pour at will wine, or
a mixture of wine and water, or 'if we want' (he says) 'to play a
joke on somebody', water. Other devices include an organ with
constant air pressure maintained in a hydraulic rescr\ oir, a pump
used for fire-fighting, a fountain worked by compressed air and a
jet-propulsion steam engine. They also include some devices con-
nected with temples and religious observance, and inevitably
our moral attitude towards Hero is closely bound up with our
interpretation of these devices and of Hero's motives in designing
them, which has been a matter of much contention among
scholars.

The late Benjamin Farrington, in his Greek Science* speaks

scathingly of those Alexandrian scientists (unnamed) who worked
for the Ptolemies, and whose science 'became the handmaid of
religion and was applied to the production of miracles in the
Serapcums and other temples of Egypt' (p. 199). He seems to im-
ply, without ever saying so, that Hero was employed in the same
disreputable line of business ('The scientific production of miracles
covers the whole period of the rise and fall of Alexandrian science',
p. 200)— but in the service of what government, or in what cult
context, he does not say. What were these Hemple miracles' ?

Of the 75 devices in the Pneumatica, 1 1 have been, or might
be, listed in that category. Two of them (I, 14 and 23) have
siphon systems by which, when water is poured into one jar, wine
comes out of another. This trick could scarcely have aroused much
*bewilderment and awe' ; for it to have any hope of qualifying as
a 'mirade', the wine would surely have to Bow from the same jar.
Four more can be effectively ruled out, because they were minia-
ture scale models — play-things of the idle rich, which could not
have had any widespread influence on the general public who are
most unlikely to have seen them. These are I, 12, II, 3 and 21,
plus I, 38 and 39, two methods of making miniature temple doors

^Pelican Books, 1953.

Copyrighted material

PRINCIPAL GREEK AND ROMAN WRITERS 203

Open when a fire is lit on a toy altar. This leaves us with five actual
^temple miracles'.

1, 1 7 is a device for making a toy trumpet play a single prolonged
note when a temple door (full-scale this time) is opened. How pro-
found a reverence this would inspire in a worshipper is doubtful,
especially if he spotted the tell-tale piece of string running from
the top of the door over a pulley. II, 9 describes a thyrsus { a wand
with imitation leaves at one end, emblem of the Greek god
Dionysus) which whistles when thrust into water. Again, we can
hardly assess the emotional effect with confidence, especially on
non-Greeks. I, 21 is a coin-in -the-slot machine to be placed at the
entrance to a temple which, in return for a five-drachma piece,
dispensed a small amount of water for ritual washing of the face
and hands. No doubt wmhippers would entirely accept the need
to purify themselves before entering a holy place, but their
thoughts on being confronted by a sort of one-armed bandit, which
stung them several days' wages for the privilege, might not have
been altogether devout. Finally, Hero explains in two places (for
the benefit of Romans unacquainted with Eg)'ptian customs) that
some Egyptian temples had bronze wheels mounted beside the
entrance, which the worshippers turned round as they passed in,
*in the belief that the bronze purifies them'. He suggests two
improvements on this device. One (I, 32) is a valve which causes
water to spout fran the hub of the wheel when it is turned, thus
streamlining the two purification rituals (by bronze and by water)
into one. The other, his ultimate essay in fiendish ingenuity (II,
32), was to mount the bronze wheel on the side of a box, on top of
which was a little stuffed bird which, when the wheel was turned,
spun round on a vertical axle and warbled. So much for the
*temple miracles', and for science in the service of religion and
oppression. Hero's other surviving works may be treated more
briefly.

(2) Automatopoietike ('the making of automata') an account of
the construction of two miniature mechanical puppet theatres.
Being designed long before the clockwork age, they are powered
by weights in the form of pistons which drop into cylinders. Before
the start of the performance, the cylinder is filled with millet or
mustard seeds, the piston resting on top of them. As the seeds run
out through a hole in the bottom of the cylinder the piston sinks
slondy down, pulling on a cord ^diich turns the main shaft and

Copyrighted material

204 ENGINEERING IN THE ANCIENT WORLD

activates what Hero calls the *plot of the play' {mythos), and wc
would call the programme. In the first device, the whole apparatus
moves forward on wheels and stops in the right position; figures
then revolve or move about, doors open and dose, fires tnim up on
tiny altars, and so on. At the end it moves back again out of d^^t.
The second device does not move as a whole but has more ^[u-
rines, performing a greater variety of movements, some of them
(rather oddly, in view of the likely audience) representing ardsans
at work. As it requiics less power, dry sand is used instead of
millet seeds in the cylinder, since it runs out more slowly, and
enables the programme to last longer.

Though these devices are of course strictly for entertainment,
they raise a number of interesting problems of design. They both
work very close to the limits of the power available, and must be
well designed and skilfully made if they are to run at alL Hero
specifies with great care tihe materials to be used. For example,
the timing of the later items in the programme is managed by
leaving various lengths of slack in the cords which operate them,
and these, therefore, must not shrink or stretch, otherwise the
sequence will go wrong. For this reason, he says, gut strings must
not be used, because they are affected by air temperature and
humidity, a fact well known to musicians, both ancient and
modern.

It has often been remarked that Hero, though he devised some
ingenious working models, never thought to apply the 'automation
principles' used in these model's to full-scaie industrial use. In
reply to this two points must be made. The power source used in
the models was very clumsy and feeble, and could not have been
made effective on a larger scale. Imagine, for instance, a weight
of 5501b (250kg) hoisted to a hdght of 13ft (4m) with a block-
and-tadde. With its rate of fall suitably controlled, it could do one
man's work (that is, develop 0.1 h.p.) for just over two mmutes,
and would then have to be hoisted up again. It is, of course, not
really a power source at all, but a device for storing energy, and it
would be much simpler and more efficient to use man-power to
work the machine directly, instead of hoisting weights. Secondly,
'automation' is a relative term. In the second device the arm of a
hgurine representing a blacksmith holding a hammer is made to
rise and fall by a sort of cam and lever arrangement similar to that
by which the windmill in PneunuUiea I, 43 pumps the oi^gan

Copyrighted material

PRINCIPAL GREEK AND ROMAN WRITERS 205

(p. 26). In order to reproduce this on a laiiger scale, a different
power source would be needed. Water-wheels may have been used

some three centuries later to work stone-saws in this way, but
Hero was apparently not acquainted with them. And e\'en if the
hydraulic hammer, worked by a water-wheel, had been developed
in antiquity as it was in the eighteenth century, it would be a Httle
misleading to call that 'automation'. The term nowadays usually
impUes the complete replacement of himian skill as well as human
physical effort, and to use a hydraulic hammer, which thumps
regularly with the same force on the same spot all the time, is by
no means an unskilled job.

(3) Mechamca, The original Greek text of this work, apart
from a few short excerpts, does not survive, but we have a
veraon in Arabic, translated from the Greek in the mid-ninth
century a.d. by a scholar whose name is usually Anglicized as
Costa ben Luka. In addition to the text, this version contains
a number of figures, possibly drawn by the translator to illustrate
his work, or derived from his Greek text, in which case they
might even be Hero's own. The treatment of perspective is odd
and sometimes confusing, and later copyists of the manuscripts
may have introduced some errors, being perhaps quite ignorant
of mechanics, but the drawing? are of great interest to the historian
of mechanical draughtsmanship, being among the earliest at-
tempts to represent such things as gears, levers and pulleys in two
dimensions.*

The treatise is divided into three books. The first (after a descrip-
tion of a geared winch, which is probably misplaced) deals with
theoretical principles — gear-ratios and their implications, the
parallelop^ram of forces, the scaling-up or down of various plane
figures, using a pantograph with two rack-and-pinion drives in a
fixed ratio to each other, and another device for scaling up or down
a 3-dimensional figure. Then pulleys are dealt with, and the gear-
ing-down effect of block-and-tackle arrangements, and in the last
11 chapters, dosely based on works of Archimedes now lost, there
is a study of the location of centres of gravity and the distribution
of loads.

*A substantial section of A. G. Drachmann's book The Mechanical

technology of Greek and Roman antiquity (Copenhagen and Wisconsin, 1963)
is devoted to this work, with tramlations of the crucial passages and
commentary (19-140).

Copyrighted material

206 ENGINEERING IN THE ANCIENT WORLD

Book II deals with the basic mechanical devices — the windlass^
the lever, the puHey^ the wedge and the screw, the last being used
either with a slider which engages the thread or with a cog-wheel
(worm-and-pinion). Each of these is investigated in turn and then,

typically, in various permutations and combinations. The mathe-
matics of worm gears — the pitch and spacing of the thicads, the
best profile for the groove and the cog-teeth, and so on — are
treated in some detail.

Book III deals with practical applications of these basic devices,
mostly in cranes and hoists, and presses. This section takes us right
into the practical mechanics of the ancient world, and is one of
the most valuable for the historian of technology. Hie descriptions
include sledges for transporting heavy stones, cranes of various
types — the single-pole jib, the shear-1^ and the derrick, presses
of various kinds, a very ingenious tool for cutting a female thread
(Hero actually uses that term, tkelu in Greek) in a block of wood,
and some devices for attaching cranes to blodcs of stone, including
the lewis bolt.f

(4) Catoptrica, the theory of mirrors. This treatise survives only
in a medieval Latin translation, in which it is ascribed to Claudius
Ptolemaens, but was almost certainly by Hero. After a brief review
of earlier theories, it discusses the phenomenon of refraction, using
the concept of a minimiun light path to explain certain effects.
Then various types of mirror are discussed, flat, concave and
convex, and some special arrangements are described for useful or
trick effects.

(5) Metrica, in 3 books, is a treatise in pure geometry on men-
suration. Book I deals with plane figures, triangle, trapeziimi,
polygon, segments of circles, ellipses, etc. Book II goes on to rectan-
gular solids, spheres and segments of spheres, cones (right and

scalene), prisms, polyhedra and others. Book III deals with methods
of dividing areas (both plane and spherical) in given ratios.

(6) Dioptra. This work deals with the theory and practice of
surveying, the opening section being taken up with a detailed
description of the surveying instrument which gives its name to
the whole work. This is a fairly sophisticated plane table, with

fA number of short extracts from the original Greek text of the
Mechanics have been preserved by Pappus of Alexandria (third/fourth
century ad), and one passage, on the geared winch, is contained in the
Greek manuscripts of the Dioptra,

Copyriyi uuj I . id^Li lai

PRINCIPAL GREEK AND ROMAN WRITERS 207

sights and water-levels, adjusted to the correct alignment by worm-
gears. It is used in conjunction with vertical poles, which have
marker discs which slide up and down them, to determine the
relative heights of various points. From Chapter 6 onwards a
series of procedures to meet various surveying problems is out-
lined ; how to take a set of sightings along a water supply route,
add up all the rises and falls from one to the next, and calculate the
total rise or fall. If there is an adequate fall, a gravity-flow system
will be feasible, but if the delivery point is higher than the source,
pumps will have to be used, and it is here that he casually mentions
several types (multi-bucket, chain, screw, drum) clearly assuming
that their design Is known to his readers. Other procedures deter-
mine the distance between two points not visible from each other,
measure range (by a sort of crude triangulation) or locate the
starting-points for a water-tunnd on opposite sides of a hill. An
apparent resemblance between the shape of the hill Hero seems
to be talking about and that of the mountain in Samos through
which a water-tunnel was cut in the sixth century B.C. ha.s led
some scholars to suggest that he had that particular work in mind;
but many hills are very like many others, and there seems little
point in choosing as an illustration an engineering problem which
had been solved many centuries before. The positions for vertical
air-shafts, required at regular intervals along such tunnels, can
also be found by the dioptra. The next problems concern land
measurement and area divisions, and then, in Chapter 34 a hodo-
meter (road-measurer) for long distances overland is described. It
is activated by a chariot wheel, and has a series of worm gears
which turn pointers from which distances up to a hundred miles or
more can be read off. For still greater distances, or distances over
sea or inaccessible land, an astronomical method of measurement
is given, and it is here that the eclipse of the moon, which has been
used to determine Hero's date, is mentioned. The last two chapters
describe a geared winch, and are clearly remnants of the Greek
text of the Mechanica, wrongly placed here in our manuscripts.

(7) Definitions. This is the opening section of a collection of
mathematical extracts made by a Byzantine scholar in the eleventh
century A.D., and Hero's authorship is generally agreed but not
certain. As the title suggests, it is a long catalogue of terms, most
of them geometrical, but some concerned with weights and meas-
ures, from which some useful information can be derived.

Copyrighted material

208 ENGINEERING IN THE ANCIENT WORLD

(8) Geometrica, This is in the form of an introduction to geo-
metry, covering some of the same groimd as the Definitions^ but in
a rather more fragmented and disorganized state.

(9) Stereomeirica, This work on solid geometry is more charac-
terisdc of Hero's style in the Dioptra. He begins with pure theory,
the mensuration of spheres, cones, pyramids, etc., but the later
problems have a less academic and more practical ring about them.
He shows how to calculate the seating capacity of a theatre from
the length of the highest and lowest rows in the auditorium.
Though the allowance of one foot width per spectator is not
exactly generous, the arithmetic is correct. He also calculates the
number of jars {amphorae) which could be stacked in the hold of
a ship of given draught, beam and length, and the number of tiles
needed to roof a building of gtven size and shape. Some of these
calculations are very rough-and-ready approximations, to serve
an immediate practical need.*

Two other exant works of Hero were, for various reasons, not
included in the Teubner edition — The Belopoeica (on catapult
construction) and the Cheiroballistra (some specifications for a
'hand catapult'.) These were edited with a commentary by E. W.
Marsden (O.U.P. 1971) and are fully discussed in Chapter 5. A
few other works now lost, are mentioned by Hero himself and by
later writers, probably the most important being a work on water-
clocks and time measurement {peri hydrion koroskopeion), the loss
of which is particularly to be r^etted.

VITRUVIUS

The Latin treatise De Architeclura, in ten books, is the only work
of its kind to survive from the Roman world. It had an immense
influence on the thinking of architects and scholars throughout the
later middle ages and Renaissance, not least on Palladio, and thus
(indirectly) on eighteenth-century English architecture.

But of its author we know remarkably little. The name Vitnivius
is a family name (i.e. in Roman terms, the middle one of three).
The work is addr^d to the Emperor Augustus, and refers to the
re-building of Rome after the civil war which ended in the battle
of Actium in 31 b.c., and so it must date from after that time. But

^MmswratUm is another scrappy oollecdon, largely covering the same
ground as (7), (8) and (9).

ed material

PRINCIPAL GREEK AND ROMAN WRITERS 209

he does not use the title 'Augustus' in addressing the Emperor, and
since that title was given to him in 27 B.a, this suggests that
Vitruvius* work dates from the four years between 31 and 27 b.cl

Julius Caesar, Augustus' adoptive father, had first given
Vitruvius some sort of recognition, but we do not know the details.
After Caesar's death (probably during the 10 years or so of civil
war which followed before Augustus eventually achieved mastery)
Vitruvius was appointed, along with tliree colleagues, to be in
charge of the construction and repair of catapults. In other words,
he was in a technical section of the army called the fabri ('engi-
neers'), who were responsible for weapon maintenance, bridge
building, transport vdiides and other such matters. The senior
officer in charge of all these operations was called praefeetus
fabrum but Vitruvius docs not claim to have held that post. He
also gives another interesting detail, that the influence of Augustus'
sister Octavia (who was married for a time to Mark Antony) had
helped him up the ladder of promotion.

Otherwise he tells us little. Tn the preface to Book II (para 4)
he says that he is not tall, and age has made his face ugly, but this
is in contrast to the youth and elegance of Dinocrates, a Greek
architect sponsored by Alexander the Great, and Vitruvius is
really suggesting that he himself has to stand on his merits, without
any kind of unfair advantage. He may possibly have held some
official position under Augustus as 'overseer of works', which
carried a salary or pension (he says he *has now no fear of poverty
for the rest of his life'), and according to Frontinus, he was respon-
sible for standardizing pipe and nozzle sizes in the public water
supply.*

The title De Architeclura should be interpreted in a rather
broader sense than the word ^architecture' normally bears. The
root meaning of the Greek word architecton is not so much *master
craftsman' as 'craft organizer', the man responsible for co-ordi-
nating and directing the work of a number of craftsmen with
various different specialized skills, and this is how Vitruvius sees
his profession.

He begins the first book with an extended account of the educa-
tion which he considers appropriate for the would-be architect,

♦The attempt (by P. Thielscher, in Pauly-Wissowa TXA. 427) to
identify him with Mamurra, Caesar's engineer officer, is not to be
regarded as successful.

Copyrighted material

210 ENGINEERING IN THE ANCIENT WORLD

and it certainly looks a foimidable programme. He must be
literate and able to express himself clearly; be must be a skilled
draughtsman, able to draw plans, elevations and perspecdve

sketches; he must be a competent mathematician, particularly
adept at geometrical constructions and arithmetic. He must also
(an interesting sideline) be a well-read and well-educated man
with an encyclopaedic knowledge of mythology and legend, so as
to be able to plan such features as pedimental sculptures or friezes
ivithout makkig factual 'howlers'. He must be a keen student of
several branches of philosophy, particularly natural philosophy
and moral philosophy (this is supposed to arm him against avarice
and corruption!); he must also understand the rudiments of
acoustics and musical theory, and must have a general knowledge
of medicine, particularly as it relates to public health. He must
also have a good grounding in the law; he must know the legal
precedents for various measures concerning drainage rights,
lighting, etc., and must be able to draw up a contract which is
clear and unambiguous, and will not gwe rise to litigation later.
Fmally, since towns and camps had to be orientated without a
magnetic compass, he must have enough basic knowledge of astro-
nomy to work out the direcdons from the sun and stars, and he is
also expected to be able to mount and calibrate sundials at various
different latitudes. All this is in addition to his knowledge of the
central skills of architectural planning, building^ design, strength
of materials and so on. The work as a whole reflects this very wide
range of diverse interests, and it is, as the author himself says, 'a
complete encyclopaedia for the architect'.

The remainder of Book I is taken up with some definitions of
basic concepts such as 'arrangement' (dispositio) and 'proportion'
{symmetria\ and with the approach to the first and basic problem
of all architecture, didce of ates. This question is discussed in
relation to several factors, the main ones being the direction of
prevailing winds, and the availability (near at hand) of building
materials.

Book II begins with a brief account (based on common-sense
and guesswork, but proved to be quite accurate by modern
archaeology) of the earliest stages in building — primitive wattle-
and-mud huts. Then, after a short excursus on the nature of matter
(one of the great pre-occupations of natural philosophy), Vitnivius

Copyrighted material

PRINCIPAL GREEK AND ROMAN WRITERS 211

lists and evaluates the main building materials — brick (sun-dried
and kiln-fired), sand, lime (calx\ and Pozzolana^ a kind of volcanic
dust found near the Bay of Naples, which was used as a light,
strong and waterproof cement. In Chapter 7 he goes on to the

principal types of stone — marble, tufa, sandstone, soapstone, etc.
Then (Chapter 8) comes an account of various techniques of wall-
building, and in the last two chapters a critical catalogue of
various types of timber.

Books III and IV deal with the siting, design and decoration of
temples, with particular emphasis on the 'orders', Ionic, Doric and
CSorinthian. Book V describes the other public buildings in a town
— forum, basilica, baths, etc In the course of his description of
the plan for a theatre, he refers to a system of resonating jars for
the enrichment of the acoustics.*

Book VI begins with an excursus on climate and its relationship
to building design, and then comes an extended account of house
design and planning, for town houses, country villas and farm
buildings. Book VII is mainly concerned with decoration, internal
and external, the preparation of stucco, colouring materials and
so on.

The last three books deal with particular problems which to us
seem peripheral to the actual business of building, but are none
the less important. Book VIII deals with water supplies and
engineering, and Book IX with astronomy and optics, and the
whole technology of time-measuring devices, sundials and water-
docks. The last book describes a number of mechanical devices —
cranes, water pumps, water wheels, catapults and other si^
engines.

FRONTINUS

Unlike Hero and Vitnivius, who were concerned with a diverse
range of subjects, Frontinus concentrated on a spedalizcd area of
technology — water supplies and engineering — but we know
rather more about him as a person, and about his career outside

the field of technology.

His full name was Sextus Julius Frontinus, so he must have
belonged (though probably in an obscure and minor branch) to

♦See J. G. Landels — 'Assisted resonance in Greek theatres', in
Greece and Rome XIV/1 (1967) 80-94.

212 ENGINEERING IN THE ANCIENT WORLD

the great aristocratic family of the Julii.* He was bom about a.d.
35., and held his first major political office in aj>. 70. He reached
the highest rank available to him under the emperor when he
became consul for the first time in aj>. 73. He was made governor
of the province of Britain for about 3— 4r years and may have been
responsible for the establishment of a legionary station at Isca
(Gaerleon). We know little of his activities for the next 20 years,
except that he composed a work, which still sur\'ives, called
Strategemaia — which is a collection of anecdotes from Greek and
Roman history illustrating the value of various tactical measures,
a sort of Field Officers' Manual. Surprisingly, it makes no
reference to his own military experiences, and is quite different
in tone and style from the later treatise on aqueducts.

In AJ>. 97 he was given responsibility for the water supply of
Rome (cura aquarum) by the emperor Nerva. ThiSy of course, is
the period o{ his life wi^ which we are mainly concerned. We
cannot say for certain how long he held the office, but he died in
Ajy. 103 or 104, so it was probably for most of the remaining years
of his life. During that period he was honoured with the Consul-
ship twice more, in February 98 and January 100, each time as a
'colleague' (so the polite fiction maintained) of the emperor Trajan.

Frontinus, in fact, presents us with two contrasting images, and
something of a problem for the social historian. On the one side
we have the patrician, with at least some blue blood in his veins,
owning villas near the sea at Formiae and Terradna, and follow-
ing the conventional career of a Roman aristocrat, via political
office and military conunand. Then, after having attained the
highest rank, when he was in his early sixties, he took on a totally
different, and apparendy much less exalted commission. It is true,
as he himself says, that the health and well-being of the whole
urban community depended on the efficient management of the
water-supply, and he adds that the office had regularly been held
by 'some of the most outstanding men in the state' {per principes
civitatis viros).

From Frontinus' own work De Aquis, particularly the Preface,
comes the contrasting image of a man who did not, as a Roman
aristocrat was conventionally supposed to do, consider the techni-

'^Xhcre sre more than 500 members of this family listed in the Pauly-
Wissowa Encyclopadie, of which Frontinus is no. 243 (VoL X p. 591).
In most other works of reference he is listed under F.

Copyrighted material

PRINCIPAL GREEK AND ROMAN WRITERS 213

cal detaik of water engineeriiig beneath his dignity. His first action
on taking over was to make a detailed personal inspection of the

entire aqueduct system (one can imagine the dismay this must
have caused among the minor officials) and to compile a short
treatise on the essential technical details, primarily for his own
use, but also for the benefit of his successors. The reasons he gives
for doing so, and for starting the work immediately on taking office,
show him as a combination of the conscientious pubHc servant and
the duvwd officer with experience of commanding men. *I have
always made it my principle', he says (preface 1-2), ^considering
it to be something of prime importance, to have a complete under-
standing of what I have taken on {nosse quod suscepi). For I do
not think there is any other surer foundation for any kind of under-
taking, or any other way of knowing what to do or what to avdid;
nor is there anything more degrading for a man of self-respect than
to have to rely on the advice of his subordinates in carrying out
the commission entrusted to him.' Of course, he says, subordinates
and advisors are a necessary' part of the organization, but they
should be its servants, not the masters to whom the senior official
in his ignorance has to keep running for advice. Tliis calls vividly
to mind a phenomenon which could occasionally be observed in
technical corps of the Army in the last war — an officer whose
technical knowledge was less that it should have been, and who
was for that reason held in very scant respect by the NGOs and
men under his command.

What kind of a post was the cur a aquarum^ and how well fitted
was Frontinus to take it on? On the technical side, he had no
special training. Such knowledge ajs he shows must have been
derived from his own reading, mainly from Greek authors who
dealt with the elementary principles. The limitations of that know-
ledge he shared with virtually all the scientists of antiquity, and it
would be unreasonable to expect a Roman administrator to be
able to solve problems in which Archimedes himself had appar-
ently no interest.

Frontinus' military e3q)erience would be of great help in pre-
paring him for the task of handling a large organization, and
dealing with the kind of large-scale operations which went on
continuously. The whole system involved about 250 miles (400km)
of conduit, most of it underground, but some on suhstracHones
(p. 38) and some on arches. Frontinus obviously visited almost

Copy I iyl uuU I Mdj-I lal

214 ENGINEERING IN THE ANCIENT WORLD

every part of this system in person, and gives his equivalent of map-
references for the location of the sources (e.g. 'The intake of the
Aqua Claudia is on a turn-off to the left from the Sublaccnsian
Way at the 38th milestone, about 300 paces along'). He gives
exact measurements of the conduits in 'paces'* even down to half
a pace in two cases. Every part of this system had to be inspected
at regular intervals, and a considerable work-force was kept per-
manently employed in maintaining it.

This work-force consisted of two contingents of slaves. One
(numbering about 240) was maintained from the public funds,
supplemented from the water-rates charged to private consumers.
The other, numbering about 460, was the personal property of
the emperor.

Frontinus was therefore in charge of a total labour force of about
700, including overseers, 'reservoir-keepers', stononasons, plas-
terers and others. At the start of his duties he had not only to see
to the renovation of various parts of the system which had fallen

into disrepair, but also to get back some members of his work
force who, as a result of bribery, had been taken off their proper
work and put onto odd jobs for private individuals. What is more,
the income from the water rates, which should have been used to
support the 'public' force, had been diverted into the private
funds of the emperor Domitian.

For all these tasks, it must have been difHcult to find anyone
more admirably suited than Frontinus. His seniority and authority
gave him the power to check corruption and raised him above any
need to involve himself in it. In his military conunands he was
used to dealing with many thousands of troops, and all the
problems of supply, finance and administration vftddi that in-
volved. But above all, his keen interest and deep sense of duty,
which made him research his job, both in the history books, the
legal records (which he quotes extensively) and 'on the ground',
gave him an understanding which must have commanded respect
in all his subordinates, and fear in those with guilty consciences.

And if he strikes us as perhaps a little ponderous, and a little
obsequious towards his emperor, two points should be remem-
bered. The formulae he uses at the start of his work were polite

•The pace {passus) was 5 Roman feet, i.e. 4 ft 10*2 in, or 1'4785 m.
The Roman mile (mille possum) was thus lGI6yds 2 ft, or 0*9185 statute
mUet (14785 km).

Copyrighted material

PRINCIPAL GREEK AND ROMAN WRITERS 215

conventions, and had no more real meaning than (say) 'I remain.
Sir, your obedient servant'. And just occasionally he shows a
touch of ironical Roman humour. He tells us (Bk II, Chapter 115)

that the official in charge of branch-pipe connections who had
allowed illegal pipes to be connecied up underground (in return
for a bribe) was known as 'a punctis\ The joke consists in giving
a high-sounding official name to an illegal activity, as one might
say in English, 'senior commissioner for water-theft in the Ministry
of Punctures'.

PLINY

His full name was Gaius Plinius Secundus, and he is known as
'Pliny the Elder*, to distinguish him £rom his nephew Gaius
Plinius CSaedfius Secundus, whose letters survive in some quantity.
Pliny the Elder belonged to the social class known as 'knights'

{equites), which meant that, though educated and well-to-do, his
family had not previously aspired to high political office. He was
born in A.D, 23, and began his career with a spell in the army, as
an officer in the cavalry in Germany from a.d. 47-57. During the
remainder of the reign of Nero (i.e. until a.d. 68) he stayed in
obscurity, probably owing to lus distrust of the Emperor, but under
Vespasian (who became Emperor in 69) he was given various
military commands and finally made the 'chief of the fleet'
(praefectus ckusis) based at Misenum on the Bay of Naples. This
was an administrative post, which carried a great deal of responsi-
bility for Mp building and maintenance, supplies, finance, etc.

This post was, indirectly and by coincidence, the cause of his
death. All his life he had a passionate interest in natural philo-
sophy, especially the more impressive and unusual phenomena of
nature. He was with the fleet at Misenum when the great erup-
tion of Vesuvius occurred in August a.d. 79., and he insisted on
putting out in a small boat to observe it from closer quarters. At
some risk to himself, he put into the shore to pick up and evacuate
some survivors, but was himself overcome by the sulphurous fumes
and died.

He was apparentiy a man of immense energy and indefatigable
zeal. In a letter to a friend (III, 5) the younger Pliny tells how his
unde used to get up in the morning before dawn, and work in-
credibly long hours. He had the alnlity to take brief *cat-naps'

during the working day, and a short siesta after his midday meal

Copyrighted material

216 ENGINEERING IN THE ANCIENT WORLD

(a light snack) enabled him to *stait the day afresh.' He was of
course very much occupM with official business, and in his early
and middle years with a legal career, but every moment of hb
leisure was spent in voracious reading of innumerable volumes by

many authors, both Greek and Latin. He had the habit (laudable

in a professional scholar, but no doubt rather annoying) of always
making notes and excerpts from everything he read (or rather,
everything as it was read to him by a slave secretary). He used
those as the basis for his extensive writings on cavalry warfare,
biography, history, oratory, and (this alone of all his works has
survived) natural philosophy. The Latin tide is NaturaUs Historian
the word historia having its older Greek meaning of 'enquiry* or
^research'.

It is in 37 books Q.e. 37 rolls of papyrus in the original manu-
script text) and contains, as he proudly boasts in the preface, *more
than 20,000 important facts, drawn from 100 principal authors'.

The first 'book' is really an index of topics and sources; the subjects
are then dealt with as follows:

Book 2 The universe (stars, planets, phenomena,

astronomy).

Books 3-6 Descriptive geography of the world.
Book 7 Human anatomy and physiology.

Books 8-11 Zoology, descriptive.

Books 12-19 Botany; descriptive accoimts of plant struc-
ture, seeds, reproduction, etc.

Books 20-27 Medical substances derived from plants, their

curative properties.

Books 28-32 Medical substances derived from animals.

Books 33-37 Mineral substances (metals, stones, etc.),

mining, metallurgy and the use of derived
substances in medicine, painting and archi-
tecture.

Pliny used a wide variety of sources in compiling his encyclo-
paedia, and he was \vont to say (again, according to his nephew)
that *no book was so bad that he could not get some benefit from
it'. This atdtude, charitable and likeable though it may be, had
unfortunate consequences. The quality of his work varies enor-
mously, from the first-rate (when he is f oUowing good sources and
even more so, when he can bring his own experience to bear) to

Copyrighted

PRINCIPAL GREEK AND ROMAN WRITERS 217

the deplorable, when he is not in a position to see the worthlessness
of his source, or has been unable to interpret it correctly. Apart
from thb variation in quality, Pliny has two characteristics which
sometimes strike the reader as regrettable. He has a great fondness

for digression and anecdote, which leads him away from his
central topic from time to time, and he also moralizes at length
on certain topics, particularly wealth and extravagance, which he
attacks with a puritanical zeal whene\ er the opportunity arises.

However, his work was of great value to the Middle Ages and
Renaissance as a storehouse of the knowledge of antiquity on a
wide range of subjects, and as such, it is of great value also to the
historian of science.

Copyrighted material

Bibliography

This 18 a very small selecdon of books and articles which may be found
useful. More extensive lists can be found in the IRstory of Ttehnologj^f and in
a number of the other works listed here.

Chapter 1

There is no general work on this range of subjects which can be firmly
recommended. R. J. Forbes deals with water-wheels and windmills in
Chapter 17 of the History of Technology, Vol II (O.U.P. 1956).
On water-wheels see also

L. A. Moritz, Grain-mills and flour in classical antiquity (O.U.P. 1958),

particularly Chapters 15-16, and Plate 14 (c).
A. W. Parsons, A Roman water-wiU in the Athenum Agora, in Ht^eria V

(1936), 70-90.

Fernand Benoit, Vusine de meunerie hydrauUque de Barbigal, in Btmu

Archeologiqtu (6), vol. XV (1940), 19^0.
£. C. Curwen, Tiu problem qf early waUr-miUs, in AntifuUy XVIII (1944)

130-46.

On the use of oxen to propel a ship, see

E. A. Thompson, A Roman reformer and inventor (O.U.P. 1952).

COiapter 2

The most detailed account of the Roman aqueduct system is
Thomas Ashby, TTu aqugduets of aneitnt Rome (O.U.P. 1935).
For some comments on Roman water engineering, see
L. Sprague de Camp, The Ancient Engineers (M.I.T. Press, 1970), especially
Chapter 6.

Chapters

On extraction of water from mines, see

R. J. Forbes, Studies in aneieni technology (E. J. Brill, Leiden 1963) Vol. VII,
Chapters.

R. £. Palmer, J^otes on some ancient mine equipment and systems^ in Transaetians
qfthe ^istituU qf Mining and MOaUmgy XXXVI (1926/7) 299-310.

Chapter 4

On the written sources, particularly Hero, see

A. G. Drachmann, The mechanical technology of Greek and Roman emHquity

(Copenhagen and Wisconsin 1963).
also

J. J. Goulton, Lifting in early Greek arekiUetme, in Journal of lUUemeShui^f
vol. XCIV (1974), 1-19.

BIBLIOGRAPHY 219

W. B. Dinsmoor, An archaeological earthquake at Olyn^iOf in Ammean Journal

of Archaeology, XLV/3 (1941), 399-427.
J. W. Shaw, A doubU-^hiooedpuU^ block/rom Cendmae, in Hespma XXXVI/4

(1967), 389-401.

Chapter 5

E. W. Marsden, Greek and Roman artillery. Vol, I Historical Development
(O.U.P. 1969), Vol. II The Technical Treatises (O.U.P. 1971).

The standard Greek texts of Hero and Pliilo are in C. Wcscher, La
PdionHiqm its Grtes, Paris 1867.

Chapter 6

J. S. Morrison and R. T. Williams, Gnek Oand Ships (G.U.P. 1968).

L. Gasscm, ^ips and seamanship in the ancient world (Princeton U.P. 1971).

On the trireme problem :

J. S. Morriscni, The Greek trireme, in MarinM^s Mirror 1941, 14ff.

Ghapter 7

On the use of oxen, see

Alison Burford, Heooy transport in atUiquityf in Economic History Review XIII

(1960), 1-18.
On horses and harness:

M. Hilzheimer, Tlie evolution of the domestic horse, in Antiquity IX (1935), 133.
R. J. E. C. Lefebm des NoSttes, Lt chaal de sdU i trovers les ages (Paris,
1931).

P^iil VIgneron, Lt duwd dans PanHfutU greahrommm (2 vob. Nancy 1968).
On the warehouses at Oslia:

H. P. Rickman, Roman granarits and start buildings (G.U.P. 1971).

Ghapter 8

G. E. R. Lloyd, Early Greek Science, Thales to Aristotle, and Greek seienu
after Aristotle (London, Chatto & Windus 1970 and 1973).

Aristotle's Mechanical Problems can most conveniently be found in the
Loeb Glassical Library under the title 'Aristotle, Minor Works', ed.
W. S. Hett.

For Archimedes' OnJIoaUng bwUts, see T. L. Heath, The works of Archimedes

(Dover, N.Y. 1958).
For the remaining texts, see notes on the next chapter.

Chapter 9

Hero's Pneumatics was first translated into English by Bennet Woodcroft in
1851. A facsimile edition, with introduction by Marie Boas Hall, was
published in 1971 by Macdonald (London) and Elsevier (New York).

A translation of his Belopoeica (on catapults) is given in Marsden's Vol. II
(see under Chapter 5).

His remaining works are available only in German translations, in the
Teubner ecfition:

Copyrighted material

220 ENGINEERING IN THE ANCIENT WORLD

Vol I. Pneumatica, Automatopoietike (ed. W. Schmidt 1899).
Vol II. Mechanica, Catoptrica (cd, \V. Schmidt and L. Nix, 1901).
Vol III. Metrica, Dioptra (ed. H. Schone, 1903).
Vol IV. Dcfinitioncs, Gcometrica (cd. J. L. Heibcrg, 1912).
Vol V. Stereometrica, Peri Metron (ed. J. L. Heibcrg, 1914).
VitniviuB* X># ArMUdma was edited and translated for the Loeb Library
in two vols, by F. Granger (Heinemann, London 1931, 1934). The
translation is not always reliable on matters other than building design
and materials.

For Frontinus, see Straiagems and Aqueducts in the Loeb Library, transL

C. E, Bennett (Heinemann, London 1925).
PHny's Natural History appeared in ten volumes in the Loeb Library,
trans 1. H. Rackham, W. H. S. Jones and D. Eichholz, between 1938
and 1962.

Copyrighted material

Index

Actium, BatUe of 134^ 165, 202

adits 15j 22

Aegean (sea) 133, Ifil

aeolipyle 2S

Aeschylus L2

agricultxire 58^ 6fi

Alexander the Great 152^ 200, 209

Alexandria 159-60, 162-3, 165, 200-1

Ammianus Marcellinus 131-2

amphora 164, 169, 203

Ampurias ilh.

Anamur 160

Anonymus De Rebus Bellicis 15
Antipater 17-18
Aqua Claudia 41^ 2M
aquarius 51-2

aqueduct 25, 40^ 49-50, 25
arch 41, 42

Archimedes 34, 42^ 54^ 59^ 95^ 97^ 106^

161. 187-93. 201, 205, 213
architrave 84, 92, IBS
arcuatio 40

Arias, P. 135n, 176n
Aristophanes 32-4, 141, 144. m
Aristotle 157^ 194-6. 201
Artemis, temple of 92
artemon L56
artesian well 26
Ashby, T. 52
Ashmole, B. 183n
askoma 142
assarium 11^ 81-2
Athcnacus 66, 153, 162-3
Athenian Agora 18, 23^, Iflft
Athens 14, 133, 145-8, 161. 170, 189
Attains 200
Augustus 208-9
Ausonius 25

axle 12, 63-74. 128. 180-2

ball (shot) 104-6, 120
Barbegal 18^22
Bath 43

bilges 58, 66, ai
Biton 97, 105-6, 200
bolt 104-6, 120. 124-5
Boon, G. G.TOn, 79n
bows 99-106, 120

bowstring 99, 102-5. 107. 114-15, 117.

123-4, 128-9, mi
braces 135, 155^ 166

brails 135, 155. 157. Ififi
Brindisi 159
Britain 32, 37, 212

bronze 47, 68, 76-7, 81-2, 112-13.

127-9, 138, 149. 162
bucket-chain 18, 20, 71-5
bucket-wheels 18, 20-1. 66-70, 22

Caerleon 212
Callixenus 152
cam 25, 22

capstan 10, 16, 68, 85-6. 124:
carchesion 94-8. 1 19
careening 162
cargo 154j 160-1, 1fi4-6
Carthage 111, 134, 145. 14a
carvel IM
catamaran 152, 162
catapults 99-132, 153. 209

repeater (/w/yAo/of) 10, 73^ 95-6. 123-
25

wedge 126

bronze-spring {chalcotonos) 122
pneumatic {aerotonos) 128-9
catinum 78, 81

Cato m

Catullus laS
cement 37, 45^ 91
Ccntenillo 62

cereals, transport of 133—1, 163-6
chains 20

link Zlr:2

strap 73, 124-5
charcoal 31-3
chariots 14, 177, 119
Charon of Magnesia 105
cheirobaUistra 120
chemistry 197-8
Chersiphron 92, lfl2
Chios 142
Cicero 177
cisium 122
Claudius 134, 163
Claudius Ptolemaeus 20L 206
cleats 64, 68, L55
Cleopatra 200
clepsydra 188-9, 192
cUnker 136, 123
coal 31-2
colluviaria 46-7
columella IBO

Cc

ul

222 ENGINEERING IN THE ANCIENT WORLD

column drums 9, 12^ 84, 89-90. 183-4
concrete 38^ 4J

conduit 3L 42^ 49-50, 58, 71^ 2M

congius Tlj 24:
connecting-rod 26
copper 122

Corinth, Isthmus of 183

Costa ben Luka 205

cranes 10^ 12-13, 84-98, 155, 20fi

crank 10-11,

creep 88^

Crete 26^ ifil

crossbow 99

Ctesibius 34, 75-6. 127-9, 190. 193,
206

cube root 120-1

cura aquarum 212-13

cylinder 26-30, 75-G. 83^ 120^ 128-9

Cyprus 160-1

dactyl 120n, 121-3
Davies, F. 79n
Demosthenes lfi4
derrick 88
Dinocrates 209
Diocletian ISO
Diodorus Siculus 66
Dionysius 123
Dionysus 203
dioptra 200-1, ?0fi-7
Dioscorides
didstra 102^, 123-5
Dolau Cothi 70n
Domitian 201, 2H
donkeys 15, 67, 170-3, 126
dowels 68, 116, 132
Drachmann, A. G. 205n
Dramont ^
drum 86=2
dry-dock 163
dryochoi 132

earthenware 37, 42-7
eddy 196-7

Egypt 10, 63, 67. 136, 170, 2QQ
Ephesus 92j 183
Eupalinos 4Q
euihytonos 118-20, 122=3
Exckias 134-5. 176n

fire-engine 58, 75, 7ft-fi0
flange ilh
fodder LZa
Forbes, R. J. 61 n
forceps 85, 89-90
force-pump 29, 75-83. 163
forestays 135, 154
frames 111-123, 130
frapping 138

Frontinus 34, 42, 48-56, 199, 209-15

fulcrum 117, 125
furnaces 31

Galen 108

gastraphetes 99, 101-4
gearbox 88, 206-7

gears 16, 19-20, 23-25. 67, 88-9, 205

Gelidonya 161

gradient 37-8, 50

granaries 171

gravity 189-90

gunwale 141. 144

hair 108, llQ-11

handspikes 10-11, 85, 89, 104, 124

halyards 135

harness 1 73-7

Harris, IL A. 181n

Haterii 12, 84, 93

heel iptema) 96, 117-18. 128. 131

hemp 1£I9

Hero of Alexandria 10, 26-31, 42-3,
47, 58, 71-2, 75-81, 88-92, 99-
105, 108-13, 115, 118-21, 130-1.
192-4, 199-208, 2ir ^

Hero's fountain 30, 42

Herodotus 14, 40

Hesiod 156

Hiero II of Syracuse 106, 161-2
Hippocrates 10, 35

Hodges, IL 61n
hodometer 202
Homer 134^, 142, 153
horn 77, IDO

horse 13-15. 170-1. 17?Uft
hushing 25

hull 136-9, 144-5, 148-9. 162. 167-8,
191

hydrostatics 34, 189-94
hypozOmata 138

iron 31, 63, 85, 92-3. 102. 111-13, 119.

124, 128, 130-1, 20a
irrigation 58-9, 65, 67, 74-5
Isidorus of Abydos 106
Isis (ship) 160-1

Josephus 165
Julius Caesar 2QD.

Kabeira 12

keel 133, 136-7, 155

keekon 136

kinetic energy 117, 186, 196

Landels, J. G. 97n, 211n
land transport 133, 170-185
lapping 29, 26
launder 2L 64, 69-70
lead 42-4, 162. 183

Copyri,,

INDEX

223

Lesbos 147

lever 25^ 194-6. 205-6
lewis 91-2
Libya HQ

Lloyd, G. E. R. 84n
lubrication 76^ Ifil
Lucian 139, mO
Lucretius 17-20
luffing 95n
lugs 80, 90, L13
Lycus (river) 12

Macedonia 148
Malea, Gape 161
Mantua 1E5
Marcellus 96
Mark Antony 209

Marsden, E. W. lOln, 105, 107. llln.

128n, m
mast 135, 140, 154-5, 158. 161. 166
mean proportional 12 1
michane 59, 62
mechanics 194-7
Megara 40
Menandcr 5H
Messina, Straits of 159
Metagcncs 18!^
meteorology 34
milHng 15-20, 22=5
mina 120n, 121, 122
mining 15, 25, 39, 58, 66, 69
Misenum 215

missile 99, 104-5. 114. 121. 123

Mitylene 146-7

Moritz, L. A. 24n

mortise 136-7

Morrison, J. S. 144

Moschion 161

mules 15, 170-7

Naupactus 149
Nero 215
Nerva 201, 212
neuron 108-9
Nonius Datus 53
nozzle 49-56, 80, 209

carports 142

oars 141-4, 151. 154. 195-fi
Odysseus 100, 135-7. 154
oU (olive) 76, LU
onager 130-2

organ 26. 75. 77, 128, 202

Ostia 134, 16L 165. 170-1

oxen 13-16. 67, 109, 170, 173-9. 183-5

Oxyrhynchus 59

Paconius 184-5

paddles 16, 20, 22, 64, 69

Palatine Anthology 12

palintonos 119=22
Palmer, R. E. 69, 2ia
Pappus 206
papyrus 1Q2
paraboloid 1^
parexeiresia 144
Pardienon 84, 126
pawls LL ni3

Payne-Gall wey, R. W. F. 132n
Pciracus 145, 148, 161
Peloponnesian War 100, 133^ 12Q
pentaspaston 85
penteconter 142, 146, 166-9
Pergamon 44, 47-8, 106. 2QQ
Pfuhl, E. 58n
Pharaohs 10

Philo of Byzantium 73, 99-100. Hi.

117, 120-31
pinion 24, 206
pipes 3L 42-50, 77-8, 209
piston 26-30, 75-6, 83, 128=9
pitch 162
Plato 160, 182
Pliny (the Younger) 215-16
Pliny (the Elder) 25, 32, 75, 92, 159.

162, 190, 199.215-17
Plutarch 96, 153, 182
pollution (water) 35, 42
Polybius 95-7. 142
Polycratcs 40
Pompeii 61
Pont du Gard 41-2
porter 170-1
protein 177-8
Ptolemy II 152

III 162-3

IV 152, 163

pulleys 10, 30, 85-9, 93-5. 155. 205=6
pumps 12, 58-83, 202
Punic Wars 134, L5i

qanat 40
quadrireme 151
quinaria 51, 56
quinquireme \hl

ram 138, 146, 149
ramming 133, 140, 149-51
ramp 10
ratchet U, 103

Reading University Project 107, 109,
U2

reservoirs 25, 48-9, 51, 67, 25
resin LLl

Rickard, T. A. 62n
Rio Tinto 12, 68-9
rocker-arm 25

Rome 34, HI, 134j 160-1. 165, 170.

177, 200-1
ropemaking 109

Copyn.

224 ENGINEERING IN THE ANCIENT WORLD

Rostovzeff, M. 61n, 177n
rotor 59-fiQ
Rouanet, G. 81n
rumen 177-8
Ruwar 2^

sails 26, 139-40, 154-162
sailing speeds 156, 1 59-6 1
St Malo as
St Paul m
Saldae 52
sambuca iiZ
Saraos 40i 202
Sanquer, R. 83n
saws 25

Schramm, General LLl
screw-pump {cochlias) 59-66
scripulus 54^ 5fi
Segovia 4J.
settling tanks 45^ 48

shafts 15, 38-9, 73-4. 8B
Shaw, J. W. 93n
shear-legs 85, 94, 96, 206
sheets 133, 135, 155, Ififi
ships 13, 58, 66, 81j 95-7, 106. 131-169,
171, 196
propulsion of 13^ I 'v-lfi
shipyards 134, 145
shrouds 155
Sidon 160-1
siege engines 105, 123
Silchester 7R-9
sinew 100-1, 106. 130-1
sinew-rope 109-13. 120. 126. 125
siparum 150.

siphon 45, 192-4, 20L=2

smelting 31, 127

Sophocles 133, 176n

Soranus lOn

Spain 25, 66, 80, 12Q

spanner 1 13-14

Sparta IQQ

sprocket 22

stability 148, mi

static electricity Ifil

steam engine 2S=3i

stem post 136, 138

steering oars 135, 139-40. 157. 160

Strabo 11

Strato 128, 201

substructio 38

Sunium IT, 142

Syracuse 95, 148, 190

tacking 158-9
talanton 120n

tendon 10ft-in
tenon m, 136-7
teredo navalis
Theophrastus L56
Thielscher, P. 209n
tholepins 141-4, 166
Thrace 148

Thucydides 100, 14^9
Tiber 134i 120
tiller 139

time measurement 188-90, 193-4-

tin 43, 122

Titus 201

torque 12, 115

torsion springs 106-23. 130

toys 2L 30, 909^

tramways 182-3

Trajan 130, 201^ 212

treadmill 11-13, 6L 65, 6L 71, 74, 82

Trier 28

trigger lOL, 123-5, m
trireme 141, 143, 145-51. 16fi-9
trispaston 85

tunnel 15, 38-9. 47, 53, 202
tympanon 26, 63, 66-7

underplate 113-15

Valverde Huelva 80=1

valves 26, 30, 75-ftl

Venafrum IB

vernier 115

Vespasian 20L 215

Virgil 166, 181, m5

Vitruvius 16-25, 34-8, 42-9, 52, 59-68.
71-9, 81, 84, 86-8, 93-5, 98, 120-
2, 131, 183-5, 199, 208-11

voussoirs 41

washers 112-15, 126, 130-1
water supply 12, 34-57

discovery of 36-7
water-wheel 16-20, 23-4, 205

undershot 16-23, 74-5

overshot 17-24

vertical-shaft 18

f>owcr output of 21=2
wheels (vehicle) 179-85. 196
Williams, C. 70n
winch H, 157

windlass 10, 85-6, 89, 94, 104-5. 131.

138, 206
\vindmill 26
worm gear 29, 206

yard 135, 140, 157-8
yoke 13, 173-4. 177. 129

Copyr